Thailand Power Sector for Data Center Investors
The thesis in one paragraph
Thailand is a state-owned single-buyer electricity market that has spent two years announcing a way out of itself without ever actually opening one. Every megawatt a data center consumes today is bought from EGAT, MEA or PEA under a regulated tariff, because the Direct PPA route around them has been approved by the National Energy Policy Council four separate times since June 2024 and has still not produced a single executed contract, a gazetted electricity Third Party Access code, or a settled wheeling charge. Meanwhile the state has moved decisively on the demand side: a separate and higher tariff class for data centers, a bond you must post to reserve grid capacity, a written sufficiency letter from the regulator you must hold before you can even apply to the Board of Investment, and a three-year shareholding lock-up on PEA-connected sites that constrains how you sell down. What actually stops projects is not any of that. It is transmission into Chonburi and Rayong. PEA's connection register held 26,045.2 MW of data center applications against 4,882.7 MW it could confirm as of June 2026, and PEA's own deputy governor put the application stack above 30,000 MW two days ago. Thailand's opportunity is real and it is regional in origin: Johor is effectively full and under a partial moratorium, Singapore rations capacity in 50 MW slices every two years, and Thai land is cheaper than either. The opportunity is also narrower than the headline pipeline suggests, because the scarce asset is not land, incentives or capital. It is a confirmed connection.
The Enhanced Single Buyer is the fact every other fact hangs off
Thailand runs what its own regulators call an Enhanced Single Buyer model, and understanding it removes about half the confusion in the market. The Electricity Generating Authority of Thailand, created by the EGAT Act B.E. 2511 (1968), generates power, owns and operates the entire high-voltage transmission network, acts as system operator, and is the sole bulk purchaser of electricity in the country. It sells that power on to two distribution monopolies with mutually exclusive franchises. The Metropolitan Electricity Authority covers Bangkok, Nonthaburi and Samut Prakan. The Provincial Electricity Authority covers the other 74 provinces. There is no wholesale spot market, no retail competition, and no unbundling of EGAT's generation arm from its transmission and system-operator roles (Tilleke & Gibbins; Norton Rose Fulbright).
The practical consequence for a data center developer is that your counterparty for firm power is a state utility, your tariff is set administratively rather than by a market, and any private generator you want to buy from must either sit inside your fence, sit inside a private estate network, or use a third-party access regime that does not yet function. Everything in the rest of this page is a variation on those three options.
The governing statute is the Energy Industry Act B.E. 2550 (2007), effective 11 December 2007, which deliberately split policy-making from regulation from operation. Section 47 makes any energy industry operation, with or without remuneration, subject to a license from the Energy Regulatory Commission, with exemptions only by Royal Decree. Section 48 pulls factory, building-control, town-planning and energy-development permitting for energy facilities into the ERC's hands, with the ERC consulting the line agencies and collecting their fees. Section 49 lets the ERC order an unlicensed operator to stop or be disconnected (Energy Industry Act, full text).
The regulator does not decide anything that matters most, and it has been half-staffed for two years
Read Section 11 of the Act and the ERC's actual mandate is narrower than its name implies. It regulates within a policy framework set by government. It announces license types, sets the rules for electricity procurement and RFP issuance, monitors selection for fairness, issues orders and administrative fines, and provides comments on the Power Development Plan, the electricity investment plan, the gas procurement plan and network expansion plans for onward submission to the Minister. It does not set procurement volumes, the fuel mix, or the plan itself.
Allocation power sits with the National Energy Policy Council, established under the National Energy Policy Council Act B.E. 2535 (1992) and chaired by the Prime Minister, currently Anutin Charnvirakul. The Energy Policy and Planning Office is its secretariat and the body that actually drafts the Power Development Plan. Below the NEPC sits the Energy Policy Administration Committee, known in Thai as กบง., which approves implementing principles. The Energy Minister as of 2026 is Akanat Promphan. When you read that "Thailand approved" something in the data center space, the question to ask is always which of those four bodies approved what, because they approve different things and only Cabinet and the Royal Gazette make an approval binding.
The ERC's statutory complement under Section 10 is seven people, one chairman plus six commissioners, appointed by the King on Cabinet recommendation for a single non-renewable six-year term, barred from being civil servants, state-enterprise staff or holders of any interest in an energy business. Four commissioners' terms ended on 30 September 2024 and the selection process for replacements was still running in July 2026, roughly a two-year vacancy. The ERC's own English board page lists four names. Secretary-General Poonpat Leesombatpiboon has publicly rejected the inference, saying flatly that "Direct PPA has not been delayed," pointing to 6,015 energy licenses and permits issued since 1 October 2024 and noting that PDP 2026 is a government-led process on which the ERC only comments (Nation Thailand). No source I have establishes how many seats are formally vacant, whether outgoing commissioners continue to serve under Section 17, or whether any quorum rule binds. Treat it as an execution risk on every date in the Direct PPA timeline rather than a proven cause of delay, and note that the same short-handed ERC is the body that must set the data center tariff, the 2026-2030 industrial tariff structure, and the wheeling charges.
A generation license is a consistency test against a plan that does not currently exist
The ERC Regulation on Energy Industry Operation License and Application Procedure B.E. 2551 (2008) creates five electricity license types (generation, transmission, distribution system, electricity selling, and electricity system operation) plus four gas types. A generation license bites at 1,000 kVA. Below that you file an online notification of exemption from being an energy business operator; at or above it you apply properly (JETRO's translation of the licensing regulation).
Article 10 is the clause that matters. The ERC must check every application for consistency with the Power Development Plan, the power business investment plan and the network development plan. That makes the PDP a hard gate on new generation rather than an indicative document. Processing runs 45 days for the Office to report to the Commission and 30 days for the Commission to decide. Article 16 requires transmission, distribution and selling licenses issued to EGAT, MEA and PEA to be issued region by region. In practice EGAT is the only holder of a transmission license; the category exists and no private party has ever obtained one.
Here is the problem. Thailand has no current-generation PDP. The legally operative plan is PDP2018 Revision 1, Cabinet-approved in October 2020 with a horizon only to 2037, whose demand forecast predates both the EV and the data center load surge. PDP 2024 was drafted to a 2024-2037 horizon, went through public hearings, and was then shelved without NEPC or Cabinet approval, attributed to overestimated demand, excessive reliance on imported gas, and misalignment with net zero by 2050 (Maverick Consulting Group). Its replacement, PDP 2026, extends the horizon to 2050 and is still a draft: the drafting sub-committee signed off on details on 17 August 2026, EPPO's public consultation is set for 8 September 2026, and promulgation is targeted within 2026 (EPPO; Nation Thailand).
On present drafts, PDP 2026 runs four scenarios all exceeding 60% clean electricity by 2050, with a base case of 65% decomposed as 42% renewables, 18% small modular reactors and 5% other new technologies. A first 300 MW SMR enters the system in 2037 under every scenario; the 2050 SMR target is reported inconsistently at roughly 9,000 MW by Nation Thailand and SolarQuarter and at 2,000-4,000 MW elsewhere, and no source reconciles them. Peak demand for 2037 is forecast at 54,192 MW low case to 57,273 MW high case, which is at or below the shelved PDP 2024's 56,133 MW despite adding data centers and EVs. EPPO is reportedly protecting an average tariff ceiling of four baht per unit across the plan period as a competitiveness constraint.
The data center assumption inside the plan is worth getting exactly right, because it is widely misquoted. EPPO's demand forecast, adopted 24 March 2026, runs three data center scenarios to 2050: a low case of 6,799 MW built from BOI-promoted projects with confirmed power, a medium case of 8,811 MW, and a high case of 19,808 MW built from every application and inquiry lodged with MEA and PEA (Thansettakij). EPPO Director-General Watthanaphong Kurowat said on 28 August 2026 that the plan sizes roughly 8,000 MW of firm state supply for data centers, with Direct PPA as the parallel route, so the eight-thousand-odd figure is a state-purchase envelope rather than the plan's view of total data center load. Anyone quoting "PDP 2026 assumes 8,800 MW" as the plan's demand estimate is quoting one of three scenarios and dropping the largest.
Until Cabinet approves PDP 2026, every new generation license is being tested under Article 10 against a plan from 2020.
Private capital has been allowed into generation since 1992, but only in three sizes and only selling to the state
Thailand opened generation to private capital through three schemes defined purely by size. Independent Power Producers above 90 MW and Small Power Producers of 10 to 90 MW sell to EGAT. Very Small Power Producers below 10 MW sell to PEA or MEA. The SPP program was endorsed by Cabinet on 17 March 1992, modeled on the US PURPA framework; the IPP program followed in 1994; the VSPP scheme began in 2002 with a 1 MW cap raised to 10 MW in 2006 (ERIA). By March 2017 EGAT had signed 156 SPP projects totaling 13,983 MW, and 981 VSPP projects totaling 5,240 MW held PPAs. Sources conflict on IPP bidding history: EGAT's own explainer gives three rounds in 1994, 2007 and 2010 totaling 16,000 MW, while another account gives 1994, 2007 and 2012. All agree there have been only three IPP solicitations and none accepted after 2003, when the system went into surplus.
The contract form matters more than the categories. IPP and SPP PPAs are capacity-plus-energy contracts with heavy minimum-take: the availability payment is owed whether or not the plant runs, and minimum take is typically 100% of capacity at peak and 65% off-peak. Critics point out that EGAT must pay fixed costs even when no electricity is generated, and the automatic fuel adjustment mechanism passes that cost straight through to ratepayers, which removes any pressure to diversify away from imported gas (Climate Finance Thailand). A July 2026 government committee resolved to set up a sub-committee to review those availability and energy payments, but explicitly deferred it until the renewable PPA reforms progress.
That surplus is not an abstraction. The ERC's own working papers put average installed capacity for 2024-2028 at 56,443 MW against average peak demand of 38,507 MW, leaving 17,936 MW of surplus, or 31.78% of installed capacity. Thailand does not have a generation shortage. It has a delivery problem and an expensive contract book.
The state overpaid for solar for fifteen years and is now clawing it back one contract at a time
From 2007 the adder regime paid a feed-in premium on top of the base tariff: THB 8.00/kWh for solar VSPPs for ten years, cut to THB 6.50/kWh in 2010, plus THB 2.50/kWh for wind, THB 2.50/kWh for municipal solid waste, and small premiums for hydro, biomass and biogas, with a southern border-province uplift on top. Those contracts carried automatic renewal, and that renewal clause is the root of what is happening in 2026. The FiT regime replaced the adder from 2014-2015 with a three-part tariff, and the current 2022-2030 procurement framework pays THB 2.1679/kWh for solar farms, THB 2.8331/kWh for solar plus battery storage, THB 3.1014/kWh for wind and THB 2.0724/kWh for biogas, on fixed 25-year terms (Watson Farley & Williams).
At its 15 July 2026 meeting, chaired by Prime Minister Anutin Charnvirakul, the NEPC approved the proposal of a committee chaired by Deputy Prime Minister Pakorn Nilprapunt to fix a definite end date for every non-firm SPP and VSPP contract carrying automatic renewal, reset purchase rates to reflect current cost with solar specified at THB 2.1579/kWh, and direct EGAT, MEA and PEA to negotiate the amendments (EPPO readout).
Three qualifications that most coverage drops. First, no wind tariff was set. NESDC Secretary-General Danucha Pichayanan, secretary to the committee, said on 8 July that THB 2.16 comes from the community solar program and applies to solar first because it has a clear reference price, with wind to be repriced separately later (Thai PBS). Second, the widely quoted 565 contracts and 3,945 MW is the committee's own tally of three buckets, not a figure in the NEPC resolution, and the tariff cut applies to only the first two of them, about 561 contracts and 3,878 MW; the remaining four contracts and 67 MW face case-by-case cancellation for missing their commercial operation deadline by more than two years. Third, it is not self-executing. Implementation runs through contract-by-contract renegotiation by the three utilities, and as of today no amendments have been reported as concluded. Industry pushback is on the record from the Thai Photovoltaic Industries Association and the Wind Energy Association, and a Chulalongkorn energy-law academic has argued the state may amend these as administrative contracts but must pay fair compensation. Also note three nearby numbers that should not be merged: THB 2.16 (the committee's reference), THB 2.1579 (the NEPC solar buy-back rate) and THB 2.1679 (the community solar FiT set the same day).
There is a second live piece of policy risk in the renewable pipeline. The 3,668 MW second renewable round opened a 2,180 MW waitlist tranche in October 2024, then the NEPC halted Phase 2 in December 2024 for an audit. Roughly 2,145 MW was awarded and about 1,500 MW remains frozen pending the new PDP. Any Thai renewable pipeline dated 2024-2026 carries this.
What a data center actually pays today, and the tariff class that has been announced but not priced
Retail tariffs are a base tariff plus a fuel adjustment charge, the Ft, reset three times a year across four-month periods running January-April, May-August and September-December. For September-December 2026 the ERC first set an average of THB 3.95/kWh excluding VAT (base THB 3.78 plus Ft of 16.23 satang), then cut it by about 6 satang to roughly THB 3.89/kWh by stripping out public street-lighting costs, about THB 18 billion a year, that had been embedded in the general tariff for more than thirty years (Nation Thailand; Bangkok Post). The 16.23 satang Ft is itself a policy choice, held roughly 25 satang below an estimated cost-reflective 41.27 satang by drawing THB 16.13 billion of clawback funds. For scale on how far that can move, Ft peaked at 93.43 satang in September-December 2022 and business users paid 154.92 satang in January-April 2023.
Industrial users pay a time-of-use structure. Type 4 Large General Service (1,000 kW and above) energy charges are reported at 4.1025 THB/kWh on-peak and 2.5849 off-peak, with demand charges of 74.14 THB/kW at 69 kV and above, 132.93 THB/kW at 12-24 kV and 210.00 THB/kW below 12 kV, all excluding Ft and VAT, with on-peak defined as 09:00-22:00 Monday to Friday. That schedule comes from a commercial solar developer's tariff explainer rather than from MEA or PEA directly, because both utility sites block automated retrieval, so confirm it against the primary schedule before it goes into a model.
Thailand is also midway through a full retail tariff redesign for the 2026-2030 regulatory period. The residential leg landed in July 2026 with the first 200 units capped at THB 3/kWh, extended to rental homes, dormitories, apartments and unregistered addresses. The industrial and commercial leg is still under review and expected to complete within 2026. Anyone underwriting Thai industrial power cost is underwriting a structure that has not been set.
Sitting on top of that, the NEPC agreed in principle on 15 July 2026 to create a separate "Type 9" user class for data centers, priced to reflect the marginal cost of imported LNG and grid reinforcement, with the surplus explicitly directed at funding household relief and public lighting. Press reporting puts the expected level at THB 5-6/kWh, roughly US$0.15-0.18, against the roughly THB 3.95 general tariff and the draft PDP's four-baht ceiling (Eco-Business). Bloomberg reported the ERC would fix the rate as early as the August billing cycle; the ERC's 12 August decision for September-December contains no data center rate. No source specifies the MW threshold that triggers the class, the split between energy and demand charge, the time-of-use treatment, or an effective date. Treat THB 5-6/kWh as a reported range and a planning assumption, not a tariff.
One thing that follows immediately from the Type 9 range: a data center paying THB 5-6/kWh has a far better case for on-site generation than a factory paying THB 3.95. The tariff decision and the captive-generation decision are the same decision.
The green tariff is the only clean-power product you can actually buy today
While Direct PPA remains unusable, the Utility Green Tariff is live, priced, and in force. It rests on an NEPC resolution of 7 November 2022 and launched on 23 January 2025.
UGT1 is the normal tariff plus a small premium, backed by renewable energy certificates from EGAT's seven existing hydropower plants, with the source unspecified to the buyer. The premium is genuinely small and it fell 37% for 2026, from 0.0594 THB/unit retail in 2025 to 0.0375 THB/unit retail in 2026 (wholesale 0.0527 to 0.0314), per EGAT's own rate announcements under ERC resolutions of 28 February 2024 and 21 January 2026 (2026 announcement). Allocation is first-come-first-served in 100 kWh blocks capped at 110% of trailing twelve-month consumption, on an auto-renewing one-year amendment to an existing EGAT PPA, administered jointly by the three utilities at ugt-thai.com. Available volume is roughly 1,300-3,500 GWh a year per BOI's briefing, or about 2,000 GWh per the portal.
UGT2 is source-specified green power drawn from two portfolios of new 2022-2030 FiT plants totaling roughly 4,850 MW, sold on ten-year contracts to large users at 100% of consumption. Its retail rates are published and in force from 1 May 2026 under an ERC resolution of 1 April 2026: Portfolio A at 4.0423 THB/kWh at high voltage, 4.1477 at medium and 4.3112 at low; Portfolio B at 4.1217, 4.2295 and 4.3969 (EGAT announcement). The figures of 4.5622 and 4.5475 THB/kWh still quoted across trade press come from BOI's July 2024 briefing where they were labeled "proposed," and are superseded.
Two observations. First, at high voltage UGT2 Portfolio A at 4.0423 THB/kWh is cheaper than the reported Type 4 on-peak energy charge plus Ft, and well below the Type 9 range. Second, uptake has been thin. The official portal reported 41 subscribers and 130,000 bundled RECs as of January 2026 against roughly 2,000 GWh/year of UGT1 supply, under 7% utilization, and its companion claim of 260,000 tCO2 avoided implies an emission factor about four times Thailand's grid factor, so at least one of those statistics is wrong. The portal also still labels UGT2 as proposed and expected in 2026 despite EGAT having published binding rates effective 1 May 2026. Treat portal status text as stale.
Direct PPA and TPA in forensic detail: approved repeatedly, still not usable
This is the file that decides whether Thailand becomes a genuinely competitive place to build AI infrastructure, so it deserves the full chronology.
The statutory basis. Third Party Access rests on Section 81 of the Energy Industry Act, which obliges a network licensee to allow other licensees and operators to connect to and use its network, in accordance with three codes (connection, utilization of facilities, and system operation) that the network owner itself writes. Those codes must not harm system security, disadvantage consumers or discriminate unjustly, and, in a clause worth reading twice, "shall not cause the licensee operating the energy network system to lose benefits or to be at a disadvantage in the competition." A self-protection clause for the incumbent is written into the statute. Sections 82 and 83 let the ERC order revisions and let users challenge codes.
The chronology. The ERC issued a Third-Party Access Framework Guideline in May 2022, directing EGAT, PEA and MEA to prepare a three-part electricity TPA Code. The NEPC then approved the pilot by resolution 1/2567 at its 167th meeting on 25 June 2024, capping Direct PPA via TPA at not more than 2,000 MW and restricting it to data center companies whose parent required renewable power, on condition of equal treatment across every country the parent invests in, large-scale investment, and no sale back into the national grid. A minority of law-firm notes date this to June 2025; the weight of sources supports 25 June 2024. Hunton expected the finalized TPA Codes "within 2024" and Service Agreements from March 2025. Neither happened. The ERC finally released draft Direct PPA criteria on 3 October 2025, with a public comment window of seven days closing 10 October, sixteen months after the policy approval (ERC hearing 580). The draft TPA charge rates were approved by the ERC at meeting 39/2568 on 22 October 2025 and consulted on from 27 October to 10 November 2025 (ERC hearing 582).
Where it stands today. On 15 July 2026 the NEPC extended Direct PPA via TPA beyond data centers to all industries seeking clean electricity and lifted the 2,000 MW ceiling. The Energy Policy Administration Committee approved the implementing principle on 3 August 2026, and the ERC moved the revised electricity-purchasing framework into the public consultation required by law. Energy Minister Akanat Promphan restated the cap removal at Thailand Focus on 26 August and again on 28 August, describing a Direct PPA market "without the previous 2,000-megawatt cap, now expanded to all industries keen on clean power" (Nation Thailand).
Two qualifications on the cap. Market participants and the BOI allocation were still working to a 2,000 MW first tranche well after 3 August: WHAUP's CEO referred on 10 August to "the electricity trading liberalization project, first phase of 2,000 MW," and Kiatnakin Phatra's 25 August utilities note models 2,800 MW across GULF, GPSC and BGRIM contributing only from 2029-2030. The accurate formulation is that the statutory ceiling is gone and 2,000 MW remains the first allocated tranche.
On the code itself, be precise about what can be asserted. The ERC's own TPA register lists approved codes only for PTT's onshore gas pipeline and PTT LNG, the most recent dated 31 March 2020. There is no electricity TPA Code for EGAT, MEA or PEA on the regulator's page (ERC TPA register). Thailand's gazetted TPA regime is for gas, and it is routinely confused with the unfinished electricity one. Say that no electricity TPA Code appears on the ERC's register and the framework is not in force, rather than asserting a Royal Gazette status nobody has verified.
Nothing has been signed. WHA Utilities and Power, the most active contender, reported Q2 2026 on 10 August with 384 MW of cumulative signed solar rooftop contracts, all booked as Private PPA behind the meter rather than grid-wheeled. Its two Direct PPA arrangements are memoranda to study a model, with APEX at 17 MW in March 2026 and Synphaet Hospital Group in June 2026. Broker headlines in June 2026 about WHAUP "winning the first Direct PPA" were speculation. On 20 July, five days after the NEPC decision, WHAUP was still publicly urging the government to issue the criteria (Nation Thailand).
Who qualifies under the October 2025 draft. The gate is narrow and worth reading in the original. A data center must hold BOI investment promotion; commit to renewable energy for 100% of demand; run a continuous IT base load of not less than 50 MW per building; file a ten-year electricity plan with annual peak MW split between Direct PPA, utility supply and total, plus a load profile; have generated no income from the project at application; hold a backup supply contract with EGAT, MEA or PEA on the utility's terms; and produce at least a non-binding memorandum or letter of intent from a producer. The project must run not less than ten years. Note a distinction English coverage generally drops: the ten-year plan is a filing requirement, but the Direct PPA entitlement is sized off the five-year peak demand plan from the pilot start date (ERC draft criteria, Thai).
The generator side is equally narrow. Renewable only, or renewable plus battery storage. Not already operating and not selling back into the national system. At least 1,000 kVA installed, or total inverter capacity for solar. No PPA still binding with the state, which disqualifies every adder-era and FiT-era fleet in the country. Feeder capacity confirmed by the utility, and an EGAT system study for anything 6 MW or above. Foreign ownership of the plant is permitted subject to the Land Code. Multiple producers may supply one data center.
There is also a use-it-or-lose-it forfeiture at clause 5.1: allocated Direct PPA volume not covered by signed grid service agreements within the specified period is deemed waived. Under the 2024 draft codes, Available Transfer Capacity is allocated first-come-first-served, on a one-generator-to-one-load basis in the initial phase, with allocation retrievable if connection does not start within four months of schedule and reducible if you take less than half your allocation.
One further conflict worth flagging. The ERC's primary draft is titled for data centers and mentions semiconductors nowhere, while trade press describes the 2,000 MW as having been allocated to data centers and semiconductor businesses. No source gives a split between the two.
The charge stack nobody has finalized is the entire economics
This is the number that decides whether an off-site Direct PPA beats simply buying UGT2, and it is the single most misreported figure in the Thai market.
Two competing proposals went to the ERC's October-November 2025 consultation. EGAT proposed a flat, distance-insensitive stack. The ERC's own consultant at Chulalongkorn's Faculty of Engineering proposed a voltage- and time-of-use-differentiated one.
| Component (THB/kWh) | EGAT proposal | ERC consultant, ≥69 kV peak | ERC consultant, MV peak | ERC consultant, LV peak |
|---|---|---|---|---|
| Wheeling (transmission + distribution) | 0.2506 | 1.1218 | 1.9110 | 2.5235 |
| System security / ancillary | 0.8870 | 0.4978 | 0.4978 | 0.4978 |
| Policy expense (Ft-linked) | 0.1727 | 0.1447 | 0.1447 | 0.1447 |
| Stack total | 1.3103 | 1.7643 | 2.5535 | 3.1660 |
Source: ERC public hearing 582 document. Totals are my arithmetic, not stated in the document.
Several things follow. The widely circulated figure of "approximately 1.07 THB/kWh on a postage-stamp basis" appears in neither proposal. It traces to a Hunton client alert of September 2024 describing an earlier draft, and is still being repeated by vendors in 2026, sometimes with an ancillary estimate of 0.10-0.20 THB/kWh that is three to nine times below either official proposal. Do not carry 1.07 in a model, even hedged.
The ancillary charge deserves scrutiny because it is not a charge for services you consume. Its largest component, Reserve and Backup at 0.2919, is derived by taking the revenue requirement of dispatchable plant (EGAT plants, IPP availability payments, imported hydro and SPP firm capacity payments, averaging THB 200,034 million a year for 2024-2028) over 222,830 GWh a year to give 0.8977 THB/kWh, then multiplying by the 31.78% surplus share of installed capacity and adding spinning-reserve fuel. Voltage regulation (0.0363) and frequency regulation (0.1697) come from consultant-derived cost shares of the same base, at 4.04% and 18.9%. In other words, a Direct PPA buyer is being asked to pay a share of Thailand's stranded overcapacity, allocated top-down. Black start is not charged in the initial phase pending EGAT data collection.
The policy expense is Ft-linked and resets every four months, covering VSPP and SPP adders, the FiT additional, and Power Development Fund contributions. The 0.1447 figure is the September-December 2025 value and is three reset cycles stale.
The stack is not the whole cost. At 69 kV and above the ERC attaches a floor: if the per-unit distribution charge collects less than the actual peak demand charge, the utility may top it up to not less than 74.14 THB/kW/month, and a flat high-load-factor data center is precisely the profile where that binds. Imbalance is settled on 15-minute intervals: positive imbalance is forfeited to the network operator without compensation, negative imbalance of 0-2% is charged at the retail tariff and above 2% at up to twice that rate. EGAT had proposed a gentler 1.05x / 1.25x / 1.50x banding; the ERC subcommittee's two-band version went to hearing. Fixed fees are immaterial by comparison: THB 10,000 connection, THB 125,000 for ATC allocation, THB 125,000 to change an approved generator, THB 10,000 to transfer contract rights, THB 120,000 a year.
Two proposals were rejected outright. EGAT asked for a gas pipeline transit charge of 0.0850 THB/kWh for running plants to absorb renewable volatility, refused because the pipeline cost elements are already inside EGAT's fuel cost and energy payments. It also asked for a network loss charge of 0.1480 THB/kWh, replaced with an obligation on the generator to inject additional electricity to make good system losses, taking transmission distance into account. Losses are therefore the only distance-sensitive element in the structure, and they are settled in energy rather than cash.
Status as of today: none of it is in force. The ERC's public-hearing register shows no 2026 consultation on TPA charges; what re-entered consultation after 3 August 2026 was the purchasing framework, not the charge schedule. ERC Deputy Secretary-General Kan Saengruang said on 28 August 2026 that the ERC is targeting completion of the Grid Code and the wheeling charge within this year, set so as not to burden other electricity users (Thansettakij). TDRI's Areeporn Asawinpongphan said the following day that if the regulations are finalized this year, actual Direct PPAs emerge next year (Nation Thailand). Any model dated 2026 should treat wheeling as a range of roughly 1.3 to 2.6 THB/kWh depending on voltage and which proposal survives, and should sensitivity-test rather than point-estimate.
What you can legally build behind your own meter today
The captive route is where Thailand is least ambiguous, and also least useful at hyperscale.
Rooftop solar of any capacity no longer needs a factory operation license. Ministerial Regulation No. 3 B.E. 2567, Cabinet-approved and gazetted 27 December 2024 and effective 28 December, removed rooftop solar from the legal definition of a factory irrespective of production capacity, where previously the trigger was 1,000 kW, and dropped the environmental safety assessment (Hunton). The exemption is explicitly limited to rooftops, terraces and parts of buildings that can be occupied. Ground-mount and floating solar remain subject to case-by-case factory-license analysis, which is the awkward part, because a data hall's roof is a trivial fraction of its load.
Inside an IEAT industrial estate the exemption does less than it looks. A solar installation still requires the IEAT land-use permit (IEAT 01/2) and a business-commencement notification. The IEAT said in early 2025 it planned to waive both for self-consumption rooftop and expected to publish the notification and procedures by mid-2025. I can find no confirmation it was ever issued. For a data center sited in an estate, which most of the EEC pipeline is, this is the most consequential open item in the captive file.
The ERC licensing line sits at 1,000 kVA regardless of who consumes the output. Below it you file an online exemption notification, confirmed in roughly 10-14 working days. At or above it you need a generation license. A 100 MW captive plant is licensable whatever it feeds. The practical clock for on-site self-consumption solar is roughly 8-14 weeks end to end, dominated by utility interconnection review at 15-20 working days for MEA and 20-30 for PEA.
Zero export is the default condition. MEA and PEA classify the standard C&I service as generate-for-own-use with no sale to the utility, and once aggregate generation is material relative to the distribution transformer, industry practice references around 15% of transformer kVA rating as the trigger, the site must install a zero-export controller or an ANSI 32 reverse power relay from the utility's approved-equipment registry. The broader rule, per the International Bar Association's survey of Thai storage regulation, is that behind-the-meter generation and storage are legal to deploy under ERC licensing but excess output cannot be sold to the grid or to third parties across grid infrastructure, which is exactly the gap the unfinished TPA Code is meant to close (IBA).
Battery storage has no standalone rulebook. It appears in Thai regulation only as an adjunct: permitted inside an eligible Direct PPA generating facility as renewables plus BESS, and paid a partial-firm FiT premium under the 2022-2030 scheme. There is no published rule on standalone or behind-the-meter BESS licensing, tariff arbitrage, or commercial discharge of stored grid energy. Separately, Thai authorities are reportedly considering mandating local storage at heavy-load facilities to manage voltage and frequency, and charging operators when demand spikes force utilities to deploy emergency generation.
A draft Act on Promotion of Solar Power Usage, put to public hearing by DEDE on 26 March 2025, would go much further, abolishing licensing for self-consumption solar entirely including grid-synchronization approval and replacing it with a 30-day notification. It remains a draft with no evidence of enactment.
Finally, one route nobody has documented publicly: whether a data center's backup diesel or gas generators may be operated commercially in Thailand for peak shaving, grid-parallel export, ancillary services or capacity payments, and what grid-parallel approval requires. I found no source addressing it, only the note that operators may be charged when utilities deploy emergency generation.
The industrial estate is the loophole that already works
Here is the structural point that most analysis of Thailand misses while waiting for Direct PPA. Supply inside a private industrial estate is legally different, because the estate power company holds its own ERC generation and distribution licenses and runs its own wires. No third-party access to a state utility network is involved, so no TPA Code is needed. B.Grimm Power delivers through its own distribution networks to over 400 industrial users on 4,764 MW of installed capacity, and the ERC's own count confirms 39 electricity distribution-system licenses issued since October 2024, so the category is live and being granted. (The causal framing here is my inference from those facts plus the TPA Guideline's scope, not a quoted legal opinion.)
The gas cogeneration SPP route is operating and growing. B.Grimm reported 4,003 GWh of electricity sales in Q2 2026, up 2.7% year on year, with sales to industrial users up 3.8%, and it connected 18.7 MW of new industrial load in H1 2026 against a 50-60 MW annual target. Its projected SPP gas price for 2026 is THB 330-350 per million BTU, and it plans to import up to six LNG cargoes.
More to the point, private supply at data center scale is already being contracted without the TPA Code, and the claim that only B.Grimm has signed is wrong. Amata B.Grimm Power, the joint venture of B.Grimm Power, Amata Corporation and Sumitomo Corporation Thailand, signed a 100 MW PPA with NTT Global Data Centers for its Bangkok 4 facility at Amata City Chonburi, announced 26 January 2026 with energization targeted for Q2 2027 (B.Grimm Power); B.Grimm already supplies NTT's Bangkok 2 and 3. RATCH signed 128 MW in the first half of 2026, comprising 60 MW through Ratchaphatthana Energy and 48 MW plus 20 MW from the Nava Nakorn plant, in which GPSC holds 30% (The Standard). B.Grimm has a further 150 MW in negotiation. Sources conflict on whether the 60 MW customer sits in Saha Union Industrial Estate, Chonburi or in an industrial park in Chachoengsao.
The other worked example is the closest thing Thailand has to a private-wire template. DayOne signed a Private User PPA with Amata B.Grimm Renewable Energy for a 22 kV floating solar facility of 42.5 MWp initial capacity inside Amata City Chonburi, delivered through the PEA grid under a tripartite structure among DayOne, AMBRE and PEA, with commercial operation expected in 2027. Note two things. It uses PEA's network with PEA as a contracting party, not third-party wheeling over it, and no source discloses whether the legal basis is a PEA back-to-back retail arrangement, an independent power supply category, or an ERC sandbox, so it cannot yet be treated as replicable. And the arithmetic is sobering: on the same campus, 42.5 MWp of solar sits against 300 MW of site power capacity, roughly 14% on nameplate and far less on energy. Behind-the-meter renewables in Thailand are an RE100 layer, not a substitute for firm supply.
The grid queue is the real gate, and the number moved this month
As of June 2026, 95 data center projects had applied to PEA for 26,045.2 MW of connection capacity. PEA could confirm 4,882.7 MW, or 18.7%, split as 3,240.7 MW firm and 1,642.0 MW conditional, with 2,768.2 MW across 18 projects already under signed PPA, leaving 21,162.5 MW unserved. The investor split reconciles exactly to the total: Chinese investors 29 projects and 9,474 MW, Thai investors 29 projects and 6,778 MW, other foreign 26 projects and 7,108.2 MW, unspecified 11 projects and 2,685 MW (Thansettakij). Of the 95, 75 projects and 20,809 MW had not started construction.
Four caveats that change how you should use this. It is PEA's register only, excluding MEA's Bangkok, Nonthaburi and Samut Prakan territory and EGAT's direct-customer channel for the largest loads. It is an application queue, not committed demand, and the regulator is openly trying to deflate it. The variant "4,482.7 MW" circulating in some coverage is a transcription error that breaks the published arithmetic; drop it. And it is already superseded: PEA Deputy Governor Pongsakorn Yutthakowit said on 28 August 2026 that applications now exceed 30,000 MW, above EPPO's roughly 8,800 MW estimate, and that PEA needs three to five years of grid preparation (Thansettakij). Energy Minister Akanat Promphan has separately put potential data center and AI demand at up to 30,000 MW. Set against Thailand's roughly 55,000 MW system peak, the ERC's framing is that BOI-promoted projects at full load would consume close to half the country's peak.
Two transmission programs answer this, and they are routinely conflated.
The near-term package is real, funded and Cabinet-approved: THB 3,000 million under EGAT's existing eastern transmission security project, mandated by the NEPC on 27 October 2025, delivering about 1,150 MW through five named works. New delivery points at Rayong 2 Substation (150 MW, completed), Phanthong Substation (250 MW, operational February 2026) and Phanthong 2 temporary substation (250 MW, operational December 2026), plus transformer expansion at Pluak Daeng (350 MW, Q1 2027) and Sattahip 1 and 2 (150 MW combined). The components sum to exactly 1,150 MW (EGAT).
The large program is not yet committed. EGAT's Transmission System Expansion Project to Accommodate Large Power Consumer, at THB 31,050 million, covers Chachoengsao, Chonburi and Rayong with new extra-high-voltage substations, transmission expansion and 5,076 MVAr of reactive compensation. Its ERC consultation document sizes it to serve the first 16 data center customers totaling 3,817.6 MW over 2027-2033, at an estimated wholesale tariff impact of +0.0139 THB/kWh averaged over project life (ERC hearing 604). The frequently quoted 2,667.6 MW is that figure net of the 1,150 MW urgent package, presented as if it were the gross. The consultation closed on 21 August 2026 and no Cabinet approval is on record as of today, so the Bangkok Post's February headline that "Egat makes B31bn investment to upgrade EEC power system" got ahead of the process. Post-2028 capacity relief in the EEC is a proposal, not a commitment.
Two new gates now sit in front of any project. From 30 March 2026, developers must obtain written confirmation from the ERC that national electricity supply is sufficient for the proposed load. And since three BOI instruments were published in the Royal Gazette on 5 June 2026, including Sor. 2/2569, the sponsor must hold that ERC confirmation letter before applying for investment promotion. Since the Direct PPA route in turn requires BOI promotion, the sequencing is circular and has to be planned around. A screening subcommittee chaired by the Energy Minister, spanning BOI, EPPO, NBTC and financial and security agencies, is expected to bring a fuller framework into effect in Q4 2026 covering electricity consumption, grid stability, water management and economic contribution.
The reservation bond is the other new cost, and its status is more fragile than most write-ups admit. The NEPC on 15 July 2026 approved in principle that large data center users must post a guarantee for grid use before the state invests in expansion, plus a water-management plan; EPPO's own release names no amount. The figure of THB 4.5 million per MW has circulated in Thai press since April 2026 as an ERC draft parameter, and BOI Secretary-General Narit Therdsteerasukdi described it on 4 August 2026 as NEPC-approved, saying a 100 MW project posts THB 450 million, refunded progressively once actual consumption reaches at least 50% of declared load and in full at 70%. Note that is a progressive refund from 50%, not "half at 50%," and note that Thai sources describe a bank guarantee rather than a cash deposit, so it should not be modeled as THB 4.5m/MW of dead cash without confirming the instrument. The ERC was still drafting the rules on 13 August 2026 and declined to comment; Bangkok Post reports the measures are expected to take effect in Q4 2026 with timing set by the Energy Policy Administration Committee. Alongside it sits a 5-7 year requirement to demonstrate commercial operation, failing which allocated capacity is revoked and reallocated to industrial operators "in the real economy."
Two further constraints on deal structure. A PEA notification published in the Royal Gazette on 4 June 2026 and effective 5 June imposes a three-year lock-up running from the actual electricity usage date, during which a change of control is prohibited unless original shareholders retain more than 50% post-transfer, with PPA novation channelled through three defined routes (common control, conversion to a public company, or amalgamation). Baker McKenzie notes the notification stipulates no express penalties, with enforcement running through PEA's discretion to treat breach as a PPA breach, and that its sharpest practical effect is on lender step-in rights and share-pledge enforcement rather than on M&A as such (Baker McKenzie). It binds PEA customers only; no MEA equivalent has surfaced, which matters because much of the Greater Bangkok load sits in MEA territory. Separately, the NBTC is examining reclassifying data center services from a Type 1 to a Type 3 telecommunications business, which under the Telecommunications Business Act would require at least 75% Thai shareholding, with a decision targeted for end-2026 after consultation. That is the single largest binary risk for a foreign sponsor and it has no equivalent in Malaysia, Indonesia, Singapore or Vietnam. A national Data Centre Business Policy Committee, created by a Prime Minister's Office regulation signed 11 August and published in the Royal Gazette on 13 August 2026, took effect 14 August and adds a further approval layer above BOI and the utilities.
Where the wires actually have room, and where the water does not
Bangkok, not the EEC, holds the larger live base. BMI put Bangkok at 145.8 MW live with a 902 MW pipeline against Chonburi at 70 MW live with a 300 MW pipeline as of Q2 2026, while Knight Frank's Data Centre Atlas 2026 gives Bangkok 122 MW at end-2025 rising to a forecast 402 MW by 2027, and a colocation vacancy rate of 23.3%. National live capacity is reported at 216 MW by BMI and around 350 MW by a Fangda Partners entry guide; the two are probably measuring carrier-neutral colocation IT load and total including enterprise and telco self-build respectively, but they are not reconcilable from the public text. Report the range.
The Ministry of Energy has publicly named the provinces with spare transmission headroom outside the EEC as Bangkok, Samut Prakan, Prachinburi, Ayutthaya and Saraburi. Set that against what brokers say about developability. Savills Thailand reports Chachoengsao constrained on both electricity and water, Chonburi similarly utility-constrained despite being the preferred location, Ayutthaya carrying flood risk with parts designated as retention areas, and approvals effectively capped around 150 MW per area on local grid concerns. Transmission headroom and site developability are different tests, and both must pass.
Roughly 70% of approved data center investment sits in the EEC, with Rayong taking about 33% and Chonburi about 32% of 2025's 36 BOI approvals. The government's proposed answer is a "Power and Water Map" to steer siting, with new capacity possibly redirected toward Mae Moh in Lampang, which has existing reservoir infrastructure.
Water is the constraint Thai industry itself ranks first. A Federation of Thai Industries survey of 160 executives put water availability at 68.8% of respondents against electricity demand at 61.3% and regulatory readiness at 61.3%. TDRI estimates a 100 MW data center consumes water equivalent to the annual use of 1.3 million people and electricity equivalent to 13 million, while creating about 50 direct jobs. EEC water demand was 658 million cubic meters in 2022 and is projected at 800 million by 2027 and 1 billion by 2036, while Rayong's waterworks already runs at about 120,000 cubic meters a day against 151,200 of capacity (Mongabay). Community disputes are documented and named, centered on Khlong Tamru subdistrict in Chonburi and Ban Chang district in Rayong, with claims of missing environmental impact assessments and absent consultation, and a People's Party MP for Chonburi called in July 2026 for a parliamentary committee to scrutinize data center approval. Water rights in Chonburi and Rayong look underpriced as a diligence item relative to power.
Connectivity is a separate geography from power, which is a point worth internalizing. Thailand's subsea landings cluster at Si Racha in Chonburi, Petchaburi, Songkhla, Satun and Chumphon, and only the Malaysia-Cambodia-Thailand system lands in Rayong. The EEC campus cluster is therefore dependent on terrestrial backhaul. The NBTC approved the Google-backed TalayLink conduit on 27 April 2026, a 300 km conduit through Thai waters able to house eleven cables, landing in the Songkhla and Satun area and routing to Australia via Christmas Island, with the stated rationale of reducing reliance on the roughly 70% of Thai international traffic currently routed terrestrially through Malaysia (Submarine Networks). Latency diversity and power headroom are in different provinces.
Who is actually building, and what it costs
The named campus set is concentrated, foreign-sponsored, and mostly inside two industrial estate groups.
| Project | Capacity | Location | Sponsor / backer |
|---|---|---|---|
| DayOne Chonburi Tech Park (CTP1) | 180 MW grid, >100 MW IT live in 2026, 300 MW site, 1 GW ambition | Amata City Chonburi | DayOne (GDS International), Coatue / Hillhouse / INA |
| True IDC EEC hyperscale campus | up to 250 MW, first phase 2027 | Chonburi (EEC) | CP Group; GIP-BlackRock partnership; Microsoft anchor |
| Beijing Haoyang / Haoyang Data Center 1 | 300 MW, THB 72.67bn | WHA ESIE 4, Rayong | Beijing Haoyang Cloud & Data |
| Digital Edge B.Grimm BKK Campus | 100 MW IT, RFS Q4 2026 | Chonburi (EEC) | Digital Edge / B.Grimm Power (BGRIM ~40%) |
| NTT BKK4 (+ BKK5 land) | 100 MW grid supply, energizing Q2 2027 | Amata City Chonburi | NTT Global Data Centers |
| Bridge Data Centres QHI01 | 134 MW BOI-approved, up to 200 MW full build | Khlong Tamru, Chonburi | Bain Capital (WinTriX) |
| GSA (GSA01 / 02 / 05) | 25.6 MW live; 38 MW + 120 MW approved | Samut Prakan, Chonburi, Rayong | GULF 40% / Singtel 35% / AIS 25% |
| STT Bangkok 1 + 2 | 22 MW live + 24 MW IT, RFS Q4 2026 | Bangkok | STT GDC / Frasers Property JV |
| Empyrion TH1 | 20 MW IT, live Q3 2027 | Bang Na, Bangkok | Seraya Partners |
| Evolution TH01 / TH02 | 12 MW sold + 40 MW; 200 MW | Bang Na; Samut Prakan | Evolution Data Centres (TH01 with Central Pattana) |
| Skyline Data Center | 200 MW | Chachoengsao | DAMAC / Edgnex |
| SUPERNAP Thailand | ~20 MW, headroom beyond 60 MW | Hemaraj Chonburi 2, Sriracha | China Mobile (acquired 18 March 2025) |
| Gorilla Technology | ~150 MW net IT | Nakhon Ratchasima | Gorilla Technology |
A caution on that table. Only a handful of these figures state whether they are IT load or gross site capacity. STT, Digital Edge and Empyrion state IT load; DayOne states both separately; NTT's 100 MW is a grid supply figure. True IDC's 250 MW, Haoyang's 300 MW, Skyline's 200 MW and Evolution's 200 MW are unlabeled. Any capacity stack built by adding these is not comparing like with like.
Hyperscaler commitments run to roughly US$7 billion from the three Western names, all of which lease or partner rather than disclose owned megawatts: AWS at more than US$5 billion with its Thailand region live since January 2025 (BMI's US$15 billion figure conflicts and likely covers a wider geography), Google at US$1 billion for a Chonburi facility and Bangkok cloud region launched January 2026, and Microsoft at over US$1 billion explicitly relying on partnerships, with True IDC the named landlord. Separately, TikTok System (Thailand) received BOI approval in May 2026 for THB 842 billion, part of a THB 913 billion data center and hosting wave in a single month (BOI). Nikkei reports the TikTok figure as US$26.2 billion against BOI's US$25 billion, a currency conversion difference rather than a figure conflict. Published 2025 approval totals conflict badly: THB 458 billion across 38 projects and 2,066 MW in one account, THB 728 billion across 36 projects in another, and US$23-23.5 billion in a third, with the latter cluster internally consistent at roughly 32 THB/USD. None is a primary BOI annual publication.
On cost, the anchor figure is Cushman & Wakefield's Asia Pacific Data Centre Construction Cost Guide 2026, published 31 March 2026, which puts Thailand at US$8.8 million per MW at the mid range, up 14.7% year on year, eighth of fourteen APAC markets between New Zealand at US$9.3m and Indonesia at US$8.3m, against Malaysia at US$9.6m, Singapore at US$14.4m and Japan at US$16.0m (DataCenterNews Asia Pacific). Read it as a construction cost, not an all-in development cost, and note it reflects 2025 build costs and predates the 2026 grid and tariff changes. Two figures often cited against it do not actually conflict: Arizton publishes US$7-11 million per MW for Southeast Asia, a range that contains US$8.8m, and BOI-approved batches imply roughly US$7.8m/MW (March 2025) and US$8.2m/MW (November 2025). Individual BOI project values scatter from about US$3.5m/MW to US$19.5m/MW, but BOI registered investment is not project cost and should not be substituted for it. Gulf's own guidance of THB 140 billion for 2,000 MW over five years implies about US$2.15m/MW, which is a corporate capital plan across joint ventures against gross capacity, not a build cost.
Thai land is the clearest cost advantage. Q1 2026 industrial land averaged up to THB 9.5 million per rai in Chonburi, THB 7.75 million in Chachoengsao and THB 7.5 million in Rayong. At 1,600 square meters to the rai that is roughly THB 5,938 per square meter, about US$180-185 at 32-33 THB/USD, against a reported Johor median of RM142 per square foot, roughly US$360 per square meter. Treat the cross-currency ranking as approximate.
Capital markets tell you what the region is paying for platforms rather than for Thai concrete. KKR and Singtel agreed on 4 February 2026 to acquire the remaining 82% of ST Telemedia Global Data Centres for S$6.6 billion at an enterprise value of S$13.8 billion, implying roughly S$6.0 million per MW of design capacity across 2.3 GW and about 42% net debt in the capital structure (STT GDC). DayOne closed a US$4.5 billion Series C on 5 June 2026, the largest single equity raise in the region's history, led by Coatue and Hillhouse with the Indonesia Investment Authority joining, and is preparing a US listing; implied valuations of about US$11 billion (from GDS's 19.9% stake mark) and US$20 billion (IPO target) both circulate and conflict. Bain Capital is selling a stake in Bridge Data Centres, which owns two Thai facilities, with reported terms shifting from up to 70% at a US$5 billion valuation in March-April 2026 to about 50% above US$4 billion in July. ARC Group puts platform assets at 25-35x EBITDA against roughly 7x for regional telcos, with the premium attributed to secured power and multi-market execution rather than to buildings, which is precisely what Thailand's power-confirmation problem makes scarce.
The Thai financing market is functioning. Digital Edge and B.Grimm signed a US$880 million green loan in May 2026 across a nine-bank syndicate including Bangkok Bank, Krungsri, KASIKORNBANK, SCB, HSBC, Mizuho, Natixis, SMBC and Standard Chartered, described as Thailand's largest data center financing. True IDC's Rayong project took a dollar loan reported inconsistently at US$530 million (Crédit Agricole CIB), US$550 million (Linklaters) and US$560 million (Octus) for a 102.6 MW campus, roughly US$5.2-5.5 million of debt per MW. UOB led a six-bank THB 28 billion green loan for DayOne's first Thai project in March 2026. True IDC was reported in July 2026 seeking about US$2 billion in three tranches, roughly US$1.8 billion in dollars plus two baht tranches, which tells you where baht funding depth currently sits.
No Thai transaction has a published valuation. GIP's stake percentage in True IDC, the equity cheque and any enterprise value are undisclosed by GIP, CP Group, Baker McKenzie and Linklaters alike. There is no Thai EV per MW comparable of any kind, and no public Thai colocation rent benchmark, so a Thai yield on cost cannot be constructed from observed data. Cushman & Wakefield's roughly 13% APAC yield on cost is regional, not Thai.
Thailand is being handed demand it cannot yet connect
The regional picture explains the pipeline better than anything internal to Thailand.
| Market | Live IT capacity | Vacancy | Pipeline | Power position |
|---|---|---|---|---|
| Johor | 1,110 MW | 0.7% | 8,542 MW announced (JLL gives 1,800 MW under construction) | RP4 data center UHV time-of-use 51.09-55.18 sen/kWh; CRESS wheeling 25 sen firm / 45 sen non-firm |
| Singapore | 1,118 MW | 4.9% | 200 MW allocated Aug 2026, next call in 18-24 months | Rationed; PUE ≥1.25, ≥50% green energy, Jurong Island only |
| Jakarta | ~340-344 MW | 20.5% | ~1.5-1.6 GW | Foreign PE holds >60% of pipeline |
| Bangkok | 122-146 MW | 23.3% | ~900 MW | Type 9 tariff pending at reported THB 5-6/kWh; no Direct PPA in force |
| Chonburi (EEC) | 70 MW | not published | ~300 MW | ~90% of national connection requests; 18+ month connection timelines |
Malaysia is the reason the pipeline is moving. Prime Minister Anwar Ibrahim told the Dewan Rakyat on 24 February 2026 that applications for data centers unrelated to high technology and AI had already been stopped, for roughly eighteen months to two years, citing electricity and water demand and the risk of pushing tariffs onto ordinary consumers (Malay Mail). Johor stopped approving Tier 1 and Tier 2 facilities and rejected about 30% of new applications in 2024. The grievance is utilization: 603 MW of actual demand against 1,276 MW of declared maximum capacity, about 47% against an 85% target. Malaysia still approved RM34.6 billion across 33 projects in Q1 2026, so the freeze is selective, and Malaysia retains one advantage Thailand does not: CRESS, launched September 2024, gives data centers a working, priced third-party renewable procurement route with a published system access charge of 25 sen/kWh firm and 45 sen/kWh non-firm, roughly 4 GW of projects announced by early 2026, and revised guidelines issued 29 December 2025.
Singapore's position is structural rationing. EDB and IMDA allocated 200 MW on 21 August 2026 under DC-CFA2, 50 MW each to Digital Realty, Equinix, Keppel Data Centres and ST Telemedia GDC, all on Jurong Island, against more than twenty proposals, with a review of whether to run another call in eighteen to twenty-four months (EDB). That is an allocation lottery for incumbents, not an origination venue.
Which leaves Thailand with cheaper land, a functioning bank market, a credible incentive regime and the region's strongest declared hyperscaler slate, against a 23.3% Bangkok vacancy rate that says the current colocation market is oversupplied, a data center tariff being set deliberately above the national average, and a connection queue where fewer than one application in five can be served. The scarce asset is a confirmed connection. Everything else is available.
What is still unresolved
The wheeling charge is the largest open number and it decides whether Direct PPA beats UGT2 at all. Two official proposals sit on the record, roughly 1.31 THB/kWh flat and roughly 1.76 THB/kWh at high-voltage peak rising to 2.55 at medium voltage, both before imbalance charges, fixed fees and the peak demand floor. Neither has been adopted, the ERC has reopened no consultation on them, and the Ft-linked policy component in both is stale.
The electricity TPA Code's three parts (Service, Connection and Operation) have never been published in final form. Everything known about ATC allocation mechanics, service agreement tenor and the one-generator-to-one-load model rests on a September 2024 law-firm summary of utility drafts that may be superseded. No capacity booking portal exists in any published form.
The Type 9 tariff has no primary-source definition. No ERC or NEPC document states the rate, the structure, the MW or load-factor trigger, or an effective date. Nor does the THB 4.5 million per MW guarantee, whose amount appears in no regulator publication and whose instrument (bank guarantee or cash) is described inconsistently.
PDP 2026's final capacity additions by technology and year, its reserve margin, its gas import trajectory and the EGAT generation quota are all unpublished ahead of the 8 September consultation. Until Cabinet approves it, Article 10 licensing has no current reference plan.
The disposition of roughly 1,500 MW of renewable capacity frozen after the December 2024 NEPC audit is unresolved, as is whether any administrative court challenge is live over the legacy PPA repricing.
Whether the IEAT ever issued the promised exemption from the land-use permit and business-commencement notification for self-consumption solar inside industrial estates is unconfirmed. So is the legal basis of the DayOne/AMBRE tripartite structure with PEA, which is the only worked example of private renewable power reaching a Thai data center over utility wires.
Nothing published establishes whether a data center's backup generators may be operated commercially, what grid-parallel approval requires, or how standalone behind-the-meter storage is licensed. And there is no MEA equivalent of PEA's connection queue, no substation-level available transfer capacity publication from any of the three utilities, and no public Thai colocation rent benchmark, which together mean grid headroom cannot be mapped below province level and Thai yields cannot be built from observed data.
What to watch, in rough order of how much it moves the answer
The ERC's publication of final wheeling and system security charges, which its deputy secretary-general says is targeted within 2026 alongside the Grid Code. That single schedule determines whether off-site Direct PPA is competitive, and the spread between the two live proposals is roughly 0.45 THB/kWh at high voltage and far more at medium.
Gazettal of the electricity TPA Code for EGAT, MEA and PEA, and the first executed Direct PPA. Until one contract exists, the route is theoretical regardless of how many councils have approved it.
EPPO's public consultation on 8 September 2026 and the subsequent NEPC and Cabinet passage of PDP 2026, including which data center scenario the final plan adopts and what EGAT generation quota it sets.
The ERC's screening framework expected in Q4 2026, which should finally fix the Type 9 rate, the guarantee amount and instrument, and the four screening dimensions.
Cabinet approval of EGAT's THB 31,050 million TSLC transmission program, and whether its 2027-2033 phasing holds. This is the only committed answer to the EEC queue beyond 2028, and it is not committed yet.
The NBTC's decision on reclassifying data centers from Type 1 to Type 3 telecommunications business, targeted for end-2026. A Thai-majority shareholding requirement would reprice every foreign-sponsored Thai campus.
Whether the legacy PPA repricing actually executes. Watch for EGAT, MEA and PEA reporting concluded amendments to the ERC and EPPO, for the flat purchase tariff EPPO and the ERC then set, for the wind rate that has not yet been fixed, and for litigation.
And the simplest indicator of all: whether the number of projects holding ERC confirmation of electricity supply moves off sixteen.
Topics
th-dc-landscape · country-th · ai-infrastructure · green-finance · tf-th-report