WholeTech Picks|WholeTechFable GuideTexas Coworking
Update May 21, 2026 Datacenters in Space & the SpaceX IPO

SpaceX files for what could be the largest IPO in history — and Bastrop is doubling.

Per yesterday’s Form S-1 and Austin Business Journal reporting: SpaceX is targeting a raise of $80 billion or more under the Nasdaq ticker SPCX, with the Bastrop County facility — where the majority of Starlink satellite kits are manufactured — set to at least double in size this year. The new solar-cell factory next door is part of the company’s stated plan to power the next generation of data centers from space.

IPO filedMay 20, 2026 Reported raise$80 B+ Nasdaq tickerSPCX Starlink FY25$11.4 B Satellites in orbit~9,600 Bastrop expansion~2× this year
Looking up

What does it mean to put a data center in space?

One reason the buildout in Central Texas matters beyond Central Texas is that the same companies driving it have begun openly arguing that AI compute will eventually outgrow Earth’s available land, water, and grid capacity. The proposed answer is to put the next generation of data centers in low-Earth orbit and power them with space-based solar — and the most concrete bet on that thesis so far is being built about thirty minutes from downtown Austin.

What a “space data center” actually is

The proposal is to put racks of compute on satellites already drawing continuous solar input above the atmosphere, networked together as a constellation, and use the cold of space — ~3 K background — as the heat sink instead of cooling towers. The claimed advantages are 24/7 solar without weather or night, passive radiative cooling without water, and unbounded siting that does not displace a town or aquifer. The hard problems are launch cost, radiation hardening, in-orbit serviceability, latency for ground users, and end-of-life debris — none of which are solved at hyperscale today.

Bastrop County · FM 1209
The supply chain on the ground, the IPO at the top
SpaceX’s solar-cell factory and Starlink components plant — the manufacturing backbone for an orbital-compute thesis, filed for IPO May 20, 2026
Updated · May 21, 2026
SpaceX IPO filingMay 20, 2026
Reported raise target$80 B+
Nasdaq tickerSPCX
Starlink satellites~9,600
Starlink FY2025 revenue$11.4 B
SpaceX FY2025 revenue$18.7 B
Existing Starlink plant550 K sqft
New solar-cell factory1.1 M sqft
Bastrop expansion this year~2× current footprint

What the new factory is for, and how it connects to the report

  • It is a supply-chain plant, not a launch site. The existing FM 1209 building manufactures the majority of Starlink satellite kits on the ground; the adjacent new factory will produce solar cells — the panels that power Starlink satellites today and, per the company’s own pitch, the panels intended to power the orbital data centers in the next generation of the constellation.
  • An IPO that confirms the scale. On May 20 SpaceX filed to go public, reportedly targeting a raise of $80 billion or more — a potential record-setting IPO, with shares slated to list on the Nasdaq under SPCX. The same filing confirms the Bastrop facility is being expanded to at least double its current size this year. Hyperscale demand for the constellation is what makes both moves bankable.
  • Starlink is now most of SpaceX’s revenue. The S-1 shows Starlink’s connectivity products generating $11.4 billion of FY2025 revenue against $18.7 billion in total SpaceX revenue — well over 60% of the company. The constellation, not launch services, is the platform on which any orbital-compute product would sit, and the platform on which the IPO is being priced.
  • One campus, multiple Musk entities. Austin Business Journal reporting describes the Bastrop site as a multi-entity campus: SpaceX manufacturing, an X (formerly Twitter) presence, The Boring Company, and the Ad Astra School share the same general footprint. Cumulative-impact analysis — the kind the report says no Texas agency currently models — would have to consider all of them at once, not facility by facility.
  • The stated long-term mission is interplanetary. The same filing frames the buildout in species-redundancy terms: “By moving beyond the only home we have ever known, we ensure species-level redundancy and that the light of consciousness will not be tied to a single planet subject to the inevitable hazards of a harsh and vast universe… We do not want humans to have the same fate as dinosaurs.” Orbital data centers and Mars settlement appear in the same document as the same long-arc thesis; the Bastrop supply chain is the near-term industrial step that funds it.
  • The civic angle remains Texan. The factory is in unincorporated Bastrop County, has applied for a state “triple-jumbo” enterprise-zone designation worth up to $3.8 million in state sales-tax refunds, and routes back through the same county-zoning, water-disclosure, NDA, and cumulative-impact gaps as every other facility on this page. The supply chain for compute that may one day not touch Earth is, for now, drawing on Texas land, Texas water, Texas air, and Texas tax base.

The thing to track

Whether or not orbital data centers ship at the scale their backers envision, the supply-chain commitment is already being made on Earth — in Bastrop, in Grimes, in the chip and solar plants being approved alongside conventional hyperscale campuses. The report’s regulatory and technological recommendations apply equally to the supply chain: the water, energy, tax, and disclosure questions are the same whether the eventual user is a server rack in Cedar Creek or a satellite at 550 km.

Sources   Austin Business Journal coverage of the SpaceX S-1 filing and Bastrop campus (May 20–21, 2026); SpaceX Form S-1 / SEC filing (May 20, 2026), including the species-redundancy excerpt quoted verbatim; Wall Street Journal reporting on the IPO raise target; GEAA April 2026 report on the orbital-compute thesis. Figures rounded; SEC filing language quoted directly.
Policy Brief · April 2026

The data-center buildout, and what it costs Texas.

A plain-language summary of the 62-page report from the Greater Edwards Aquifer Alliance on how AI and crypto-mining facilities are reshaping the state’s grid, water supply, public budgets, and the air around them.

By the numbers

Texas is now #2 in the U.S. for data centers — and growing fast.

The San Antonio–Austin corridor saw a four-fold increase in construction between 2023 and 2025 and is ranked first among global emerging markets. Over 400 facilities are operating across 25 Texas markets, with hyperscale AI campuses and crypto mines driving the next wave.

494,091ac-ft
Projected Texas data-center water use by 2030 — ~3% of total state water use.
70%
Forecast statewide electricity-demand increase by 2031, driven largely by data centers.
$9billion
Projected state tax-exemption cost 2025–2030 (not counting local abatements).
225
ERCOT large-load interconnection requests in 2025 alone — vs. 152 in the prior two years combined.
1,300
Additional U.S. premature deaths/year by 2030 from data-center air pollution.
100
Permanent jobs at OpenAI’s Stargate campus in Abilene — despite 1,500 construction workers.

The thesis

AI and data centers aren’t going away — but Texas’s regulatory framework was built for a slower, smaller-load industrial economy. Without guardrails on transparency, cost allocation, water and energy use, and local control, the buildout will raise costs for ordinary Texans, deplete water supplies, and harm public health — disproportionately in rural, unincorporated, and drought-stressed communities.

What’s at stake

Four pressures, one regulatory gap.

Energy

An overwhelmed grid

Data centers already use enough power to supply over half of Texas homes. ERCOT was designed to handle 40–50 interconnection requests; it received 225 in 2025.

  • 73% of 2025 requests were data centers, 10% crypto
  • Tech firms now building private gas plants on-site
  • Crypto alone costs Texans ~$1.8B/year in extra bills
💧
Water

A planning blind spot

Texas’s State Water Plan won’t reflect data-center use until at least 2032, worsening an already-projected 5M acre-feet shortfall by 2070. Two San Antonio facilities used 463M gallons during the 2023–24 drought.

  • 75–83% of demand is indirect — power generation
  • Wastewater can carry PFAS, biocides, heavy metals
  • Companies often request more than they need
🏭
Public health

Air, noise, heat

Fossil-fueled generation creates particulate matter and NO₂. Cooling systems and generators emit constant low-frequency noise. Data centers create local heat islands averaging +3.6°F, sometimes up to +16°F.

  • ~600,000 added asthma cases/year by 2030
  • $20B+ in annual U.S. public-health costs
  • Many facilities exceed 80 dB — EPA caps safe at 70
🏢
Public money

Tax cost vs. local benefit

The state’s data-center sales-tax exemption cost $1.016B in FY2025 — nearly 8× the original projection. Counties have no zoning authority, leaving rural areas exposed.

  • Often only tens of permanent jobs per facility
  • Rockdale crypto mine: promised 350, delivered 14
  • NDAs routinely shield project details from the public
Electricity is economic. Water is existential. — Texas Senate staffer, quoted in the GEAA report
Why the rules can’t keep up

The regulatory gaps.

The state has no mandatory reporting for water or energy use by data centers. Counties have no zoning authority. Cities can be stripped of jurisdiction through de-annexation. Non-disclosure agreements bar elected officials from telling constituents what’s coming.

No state water-use disclosure No state energy-use disclosure No county zoning authority No data-center water rate class No cumulative-impact review No public NDA limits No mandatory decommissioning bonds Cooling tech not regulated
The proposed fix

Two tracks of guardrails.

The GEAA report’s recommendations split into policies the state and locals should adopt — and technologies operators should adopt regardless of what the legislature does.

🏛 Regulatory

State, county, municipal & utility action
  • Mandate water & energy reporting to TWDB, PUC, and ERCOT — publicly accessible.
  • Create a new rate class for data centers, for both electricity and water.
  • Require upfront infrastructure investment; prohibit cost shifts to residential customers.
  • Include data-center demand in the State Water Plan now, not 2032.
  • Expand groundwater conservation district authority statewide.
  • Eliminate or sharply limit tax exemptions; tie any remaining to strict job and water/energy terms.
  • Ban new crypto mines or restrict their fossil-fuel generators and drought-stage operations.
  • Give counties land-use, moratorium, and noise-regulation authority.
  • Strictly limit NDAs with government bodies; require public hearings.
  • Cap noise pollution at EPA limits — 60 dB for long-term exposure.
  • Require renewables + batteries; phase out diesel backup.
  • Encourage community benefit agreements with metering, fines, and decommissioning bonds.

🧬 Technological

For operators, regardless of policy action
  • Liquid immersion cooling — up to 50% energy and 91% water savings vs. air cooling.
  • Rainwater harvesting on the large flat roofs typical of data centers.
  • Atmospheric water harvesting — pulls vapor from the air; works well in humid climates.
  • Recycled produced water from oilfields — Texas could have 168B gal/yr available.
  • Recycled municipal wastewater — already used by Amazon and Google in many facilities.
  • Brackish groundwater with desalination — 31 production zones designated in Texas.
  • On-site solar plus battery storage — quickest to deploy, lowest water use.
  • Brownfield & abandoned-site reuse — existing infrastructure, easier permitting.
  • Underground siting where geology allows (not karst regions like the Edwards Aquifer).
  • Treat discharge to drinking-water quality before release to local systems.
Technologies in depth

How much less harmful can it get?

The short answer is: a lot, but only by stacking five or six choices that have to be made early — at siting, at design, and during procurement. None of the items below is a silver bullet, several work in some climates and not others, and a few are still pre-commercial in the U.S. This section deepens the technology column of the report’s “two tracks of guardrails,” organized by the resource pressure each set of tools is trying to ease.

Pressure · Water

Cooling that uses less water — or none

Most of the water a data center consumes goes to cooling: either directly inside the building (evaporative towers, adiabatic pads) or indirectly through the thermal-electric power plants on the other end of the wire. Replacing evaporation with closed-loop heat transfer is the single largest lever an operator can pull on water.

  • Free-air (“economizer”) cooling. Uses outdoor air directly through the racks when temperature and humidity allow. Cuts both energy and water dramatically in cool climates.Texas penalty: hot, humid summers in Central Texas leave only a few hundred hours per year of useable economizer time.
  • Adiabatic and evaporative make-up. A fine spray pre-cools incoming air. Far less energy than mechanical chillers, but still consumes water — often the dominant on-site draw at older campuses.Where water is scarce or expensive, the tradeoff inverts: cheaper to add fan and chiller energy than to evaporate aquifer water.
  • Dry coolers and sealed glycol loops. Move heat-rejection to fin-and-fan coolers; on-site water consumption falls to near zero. Pays a fan-energy and footprint penalty in summer.Best paired with cooler nighttime “free” rejection windows and high-inlet-temperature rated equipment.
  • Direct-to-chip liquid cooling (cold plates). Coolant runs through microchannels bolted to the CPU or GPU. Lets the loop run warm enough to skip chillers entirely. Standard on most new AI training racks.Requires plumbing back to the building, but cuts cooling energy 30–50% with on-site water use approaching zero.
  • Single-phase immersion. Whole servers submerged in a dielectric fluid that is pumped through a heat exchanger. Eliminates fans, allows extreme rack density, and decouples cooling from the local water supply.Mature in crypto and HPC; spreading into mainstream hyperscale design.
  • Two-phase immersion. The dielectric fluid boils at chip temperature; the vapor condenses on an overhead coil. Highest heat-density option known.Newest of the cooling options; fluid supply chain remains small, and some fluorochemical coolants face restriction under emerging PFAS rules.
  • Rear-door heat exchangers. Chilled-water coils mounted on the back of standard racks. Captures most of the rack’s heat into a liquid loop with no plumbing changes inside the server.Strong retrofit option for older facilities that can’t justify a full immersion rebuild.
  • Closed-loop, zero-blowdown chilled-water systems. The same water cycles indefinitely, treated on-site, with discharge only on long maintenance intervals.Pairs with any of the above to keep the once-through draw near zero.
Texas fitHot, often humid Central Texas penalizes the air- and evaporation-based savings; closed-loop liquid options — cold plate, immersion, rear-door — carry most of the water savings here, especially when paired with warm-water loop design that runs above 32°C.
Pressure · Power & grid

Cleaner, more flexible electricity

A data center can change two things about its electricity: what kind it consumes (carbon intensity) and when it consumes it (flexibility). The second is what ERCOT increasingly values, because Texas already has more curtailed wind and solar than most U.S. grids combined.

  • On-site solar plus battery storage (BESS). Fastest to deploy, near-zero water; can shave peak demand and replace diesel for backup.Footprint is real: roof and adjacent land typically cover 10–25% of a hyperscale load, not 100%. Best treated as one layer of a stack.
  • Hydrogen fuel cells for backup. Replace diesel generators with cleaner reserves; multi-day demonstrations have run successfully in pilot data centers.Today’s hydrogen supply chain is mostly steam-methane-reformed; the carbon advantage depends on green-hydrogen capacity scaling up.
  • 24/7 carbon-free power matching. The operator buys clean generation on an hourly basis, not just annually. Closes the gap that annual REC matching opens when a campus runs on gas at 3 a.m. and counts wind from noon.Several hyperscalers now publish hourly carbon-matching reports; none is fully matched yet.
  • Long-term “additionality” PPAs. Power-purchase agreements tied to new wind, solar, or storage projects that would not otherwise have been built.Distinct from buying unbundled RECs; the additionality test is what differentiates real grid impact from accounting.
  • Grid-interactive demand response. The data center curtails or shifts load in response to grid signals: price, frequency, or operator instructions.ERCOT’s 4CP program is the legacy mechanism; SB 6 rulemaking is shaping the next generation of curtailment protocols specifically for large loads.
  • Workloads that follow clean power. Long-running training jobs migrate between regions and hours according to renewable supply. Peer-reviewed implementations exist; inference workloads are harder to defer.Pure-play AI campuses with synchronous training schedules have less room to shift than mixed workloads.
  • Heat pumps tied to nearby thermal demand. Where a district-heating loop, industrial steam customer, greenhouse, or municipal pool exists nearby, waste heat can offset gas-fired heating.Common in Scandinavia and the Netherlands; rare in Texas, whose climate has limited use for space heating.
  • Small modular reactors (SMRs) and nuclear PPAs. Increasingly discussed for hyperscale baseload.No commercial SMR is yet operating in the U.S.; Texas has not licensed one. Realistic operating dates remain in the early 2030s at earliest.
Texas fitERCOT’s grid is unusually suited to load flexibility — the state leads the country in curtailed renewables — and data centers are among the few large loads that could absorb those curtailments without storage. The harder problem is regulatory: SB 6 rulemaking is still defining how curtailment will be priced and enforced.
Pressure · Public health

Cleaner backup, quieter rooms, less local heat

Most of the air, noise, and heat-island harms the report flags trace to two physical realities: diesel or gas combustion on-site for backup and peaking, and the building exhausting waste heat into the surrounding air. Both are addressable with technology already on the shelf.

  • BESS plus fuel-cell backup, not diesel. Eliminates the PM₂.₅ and NO₂ load from monthly generator tests and prolonged outage runs.Diesel testing alone can dominate a campus’s annual emissions inventory in residential-adjacent locations.
  • Selective catalytic reduction on any gas backup. Knocks out 80–90% of NO₂ from natural-gas reciprocating engines or turbines.Standard practice on industrial boilers; uncommon on data-center backup, where it is treated as “rarely used” equipment.
  • Sealed, insulated mechanical yards. Concrete enclosures, low-frequency baffles, and resilient mounts can keep fence-line noise under 60 dB — the EPA long-term residential limit — instead of the 70–90 dB documented at many existing campuses.
  • Variable-speed fans and pumps. Reduce the continuous low-frequency rumble that drives chronic neighbor complaints, especially at night. Slower fans run more efficiently as a side effect.
  • Liquid cooling as noise reduction. Removing forced-air cooling towers and most rooftop chillers is itself one of the largest single sources of campus noise reduction.A force multiplier on the water and energy benefits in the section above.
  • Heat-reuse offtake. Every BTU of waste heat piped to a district loop, greenhouse, or industrial customer is one that does not vent into the surrounding microclimate — the source of the +3.6 to +16°F neighborhood heat islands documented in the report.In Texas the practical offtake is greenhouse and aquaculture warming rather than space heating.
Texas fitThe mitigation menu is mature and well-priced; the binding constraint is that Texas counties have no zoning or noise authority to require any of it. The report’s call for county-level noise caps is what would unlock the menu.
Pressure · Compute itself

Doing less work for the same answer

Every watt the chip does not draw is a watt that does not have to be cooled, generated, transmitted, or supplied with water upstream. Efficiency gains at the silicon, system, and workload layers compound through the rest of the stack — a 30% chip-side gain stacked on a 30% cooling-side gain stacked on a 30% grid-intensity gain is, very roughly, two-thirds less harm per unit of useful work.

  • Right-sizing AI models. Distillation, quantization, mixed precision, and Mixture-of-Experts architectures often deliver the same task accuracy at a fraction of the inference energy.Inference accounts for the majority of total energy use over an AI model’s lifetime, so right-sizing pays out on every query, not just at training.
  • Accelerator sharing and multi-tenancy. Modern accelerators partition cleanly across many small jobs; the alternative is $40k cards spending most of their life idle.Crucial for inference workloads where each request uses only a sliver of the card’s capacity.
  • Data Processing Units (DPUs). Offload networking, storage, and security from general-purpose CPUs, freeing host cycles for actual workloads.Typical 5–15% efficiency improvement on the host with no software changes.
  • Dynamic power capping and deep-sleep states. Software ceilings on per-rack draw during grid stress; deep sleep on chronically idle hosts.Usually 5–15% savings with no user-visible impact, and a natural fit with grid-interactive demand response.
  • Warm-water liquid operation. Letting chips run hotter (45–60°C return water) shrinks the energy spent fighting heat without reducing throughput.Enabled by direct-to-chip cooling; requires hardware vendors to qualify for higher inlet temperatures.
  • Sustainable software-engineering practices. Caching, batch scheduling, and amortizing model calls across users routinely cut total inference load by 30–70% with no quality loss.
Texas fitThese are the gains that do not require a single new pipe, permit, or substation. The catch is they are invisible to the public and to regulators: no current obligation to report them, and no incentive structure to reward operators that go above the baseline.
Pressure · Lifecycle

Build smarter, retire cleanly

A campus’s resource footprint is partly cast in concrete on day one and partly determined by what happens at the end of a 15- to 25-year useful life. Both ends are technological choices.

  • Brownfield and substation-adjacent siting. Reuses graded land, existing power and water tie-ins, and avoids opening new corridors through farms or aquifer recharge zones. Often cuts construction time by a year.A natural fit for retired power plants, defunct refineries, and decommissioned industrial sites — Texas has hundreds.
  • Modular and prefabricated construction. Factory-built power and cooling modules cut embodied carbon in steel and concrete (together roughly 11% of global CO₂) and shorten the disruptive construction period for neighbors.Also tightens schedule risk — valuable when grid interconnection windows are short.
  • Climate- and demand-aware siting. Co-locate with natural cooling, waste-heat customers, or curtailment-rich grid nodes rather than chasing tax incentives alone.In Texas, paired greenhouse use is the practical analogue to the Scandinavian district-heat cases.
  • Avoid karst and recharge zones. Keep new construction off the Edwards and Trinity aquifers’ contributing zones; the report specifically excludes underground siting in these geologies for the same reason.
  • Hardware refurbishment and resale. Decommissioned servers, accelerators, and storage are routinely re-flashed, re-warrantied, and resold by hyperscalers; the same playbook is rare at smaller colocation campuses.Extends useful life by 3–7 years and recovers rare-earth content.
  • Decommissioning bonds. Money set aside up-front to demolish, remediate, and re-vegetate a campus if the operator leaves — the way some early Texas crypto sites were left.Modeled on long-standing nuclear and mining-decommissioning rules.
  • Discharge treated to drinking-water quality. Any blowdown or condenser discharge is treated before release.The report’s standard; raises the floor on water reuse rather than relying on receiving systems to clean up downstream.
Texas fitTexas already has federal Brownfields program tie-ins and an existing nuclear-decommissioning bond model that could be adapted. The harder lift is political: requiring it of a new sector that has been growing without that obligation.

The pattern

The engineering exists, much of it is mature, and most of it is cheaper than the alternatives within a 7- to 10-year horizon. What is missing is not the technology — it is the regulatory or contractual requirement to use any of it. The report’s two-track recommendation works because, absent the first track, the second is selected only on the projects where it happens to be the cheapest path. The open Texas question is whether SB 6 rulemaking and the 90th Legislature’s interim studies turn this menu into a baseline expectation or leave it on the “nice to have” shelf.

Where things stand

What Texas has already done.

SB 1929 · 2023
Requires bitcoin miners >75 MW to register with the PUC — filings have not been made public.
SB 6 · 2025
Shifts grid-connection costs to large loads, requires backup-generation disclosure, and adds curtailment protocols. Rulemaking ongoing.
FY 2026 budget
Directs PUC and TWDB to survey data centers and crypto mines on water use — voluntary, and only for operating facilities.
90th Legislature · 2027
Both chambers have pledged interim studies on electricity, water, transmission, tax incentives, and ratepayer protections.
Source   “Data Centers in Texas: A Review and Call for Innovation and Regulation”
Rachel Hanes, Policy Director, Greater Edwards Aquifer Alliance — April 2026 (62 pp.)
Case in point

What this looks like east of Austin.

The clearest test of the report’s warnings is happening about 25 miles east of downtown, in a stretch of unincorporated Bastrop County around Cedar Creek. Hyperscalers are buying land, a 1,170 MW gas peaker plant is seeking school-district tax breaks, and county commissioners — without zoning authority — have already urged TCEQ to approve the air permits. Three concurrent projects, on or next to the same site, mirror nearly every concern flagged in the GEAA report.

Bastrop County · Cedar Creek
The Cedar Creek cluster
A 1,170 MW gas peaker sited beside ~1,500 acres of planned hyperscale data centers
Active · May 2026
Amazon (AWS)1,229 acres
EdgeConneX117 + 180 ac
MPH Peakers (gas)1,170 MW
Existing gas plant598 MW
EdgeConneX plan$1.4B / 2.8M sqft
Peaker cost · jobs$371M / 7 perm

How it matches the report

  • Fossil-fuel pivot, driven by data centers. Gas generation passed wind in the Texas grid’s interconnection queue for the first time since January 2016, largely because hyperscalers want firm, dispatchable power. The Cedar Creek peaker is the local version of that statewide shift.
  • Tax incentives for an industry that doesn’t need them. Hull Street Energy is seeking JETI abatements from Bastrop ISD — the Chapter 313 successor the GEAA report calls on the legislature to eliminate or sharply pare back. Amazon-scale buyers, meanwhile, are routinely eligible for the state’s broader data-center sales-tax exemption.
  • Few permanent jobs for major capital. The peaker will create 7 permanent positions on a $371M investment; data-center campuses typically generate low-three-digit permanent headcounts even at $1B+ scale. The “lowest jobs-per-square-foot of any major industry” pattern flagged in the report.
  • Counties can’t zone. Bastrop County commissioners’ only formal move was a non-binding resolution urging TCEQ to grant the air permits — precisely the limit on county authority the report identifies as a legislative gap.
  • Land assembled through shell entities, with NDAs and “due diligence” labels. Amazon’s 1,229-acre buy was assembled across dozens of parcels by “Amazon Data Services Inc.” on land previously slated for the Creekside master-planned community; the company has labeled the purchase as “due diligence.” The transparency gap the report describes, in miniature.
  • Cumulative impacts no one is modeling. Three adjacent projects pulling power and (eventually) cooling water at the same time — the kind of cumulative load that doesn’t appear in the State Water Plan and isn’t reflected in ERCOT’s historical demand forecasts.
Bastrop County · FM 1209
The supply-chain file: SpaceX
A 1.1 M-sqft solar-cell factory pitched explicitly to power orbital AI data centers
Active · May 2026
New solar-cell factory1.1M sqft
Existing Starlink plant550K sqft
Combined buildout~1.6M sqft
Musk acreage in county~700 acres
May 26 hearing$3.8M sales-tax refund
Terafab (Grimes Co.)up to $119B

What it adds to the report’s themes

  • Orbital data centers go from rhetoric to construction line item. The GEAA report quotes AI leaders worrying that “data centers will eventually require more energy and land than are available on Earth.” SpaceX is now building the solar-manufacturing capacity specifically to put compute in low-Earth orbit — the first concrete supply-chain commitment to that thesis.
  • Chip fabs are part of the water story too. The report notes that semiconductor manufacturing for data centers is expected to double its water demand industry-wide by 2035. Musk’s proposed Terafab, separately under consideration in Grimes County at $55–$119B, is exactly the supply-chain facility that compounds the load on Texas aquifers and the grid.
  • Another flavor of tax incentive. Beyond JETI school-district abatements and the state data-center sales-tax exemption, SpaceX is seeking a “triple jumbo” enterprise-zone designation worth up to $3.8M in state sales-tax refunds. Bastrop ISD, the county, and the state are stacking incentives on adjacent facilities — against the same revenue base.
  • Transparency below even the NDA floor. Key project details surface through a director of solar production’s social-media posts and public-records requests, not formal disclosures. Officials confirm filings only when asked. The report flags NDAs; this is a step beyond.
  • A cumulative environmental record already on file. SpaceX has racked up TCEQ fines over water and wastewater impacts to the Colorado River, and a separate “aluminum melting and holding” air permit is pending at the same address — precisely the kind of cumulative-permit pressure the report says no Texas agency currently models.

Wider Austin-area context

The Austin Business Journal has identified more than 30 data-center projects in the greater Austin area in recent years, with a combined $50 billion+ in capital investment — likely an undercount, since corporate buyers regularly use shell entities and label early-stage acquisitions as “due diligence.” Bastrop County is the current epicenter, but Hays, Williamson, and Travis counties are seeing similar pressure. Add the supply-chain footprint — solar-cell fabs, chip plants — and the local resource calculus shifts further.

Sources   Austin Business Journal · May 2026: Amazon Bastrop land purchase (1,229 ac, confirmed May 13); Hull Street Energy / MPH Bastrop Peakers JETI application (filed May 14); EdgeConneX Cedar Creek campus; SpaceX solar-cell factory & enterprise-zone application; Terafab proposal (Grimes County).
The wider region

Central Texas, county by county.

Bastrop County is the current epicenter of the buildout, but the pattern reaches across at least seven Central Texas counties. Below is a working inventory of publicly reported projects and supply-chain facilities as of May 2026, drawn from Austin Business Journal coverage, county commissioners’ court records, ERCOT large-load filings, and zoning hearings. Where details are routed through shell entities or labeled “due diligence,” the line items below reflect what has been publicly attested rather than what the buyer privately intends.

Bastrop County
Current epicenter. Amazon (1,229 ac in Cedar Creek), EdgeConneX (117 + 180 ac), MPH Bastrop Peakers (1,170 MW gas), an existing 598 MW gas plant, and SpaceX’s combined Starlink + solar-cell complex on FM 1209 — with the SpaceX site itself slated to at least double in 2026. JETI applications pending; SpaceX’s enterprise-zone refund up for hearing May 26.
Travis County
Established colo base, new hyperscale interest. Long-running campuses operated by Data Foundry, CyrusOne, Stream Data Centers, and others — mostly in the Northwest Travis and MetCenter corridors. Additional acquisitions reported in the far-northwest and southeast quadrants in 2025–26, several routed through holding entities.
Williamson County
Round Rock, Hutto, Taylor. Stream Data Centers’ Round Rock campus operating; T5, Tract, and other named operators have publicly disclosed land assemblies in Hutto and Taylor. Samsung’s Taylor fab — though semiconductor, not data center — is itself a dominant AI supply-chain load.
Hays County
Aquifer-sensitive. Sits over both the Edwards and Trinity recharge zones; the report’s strongest geological objections apply here. Several smaller projects under quiet diligence; commissioner-court agendas have begun asking for setback and water-disclosure ordinances.
Comal County
I-35 corridor pressure. Adjacent to the same Edwards Aquifer recharge zone as Hays. Speculative campus filings have surfaced; municipal utility districts in the Schertz / Selma band have begun fielding hyperscale inquiries.
Bell County (Temple, Belton)
Confirmed Meta campus. Meta’s ~$800 M Temple data center is the largest publicly named Central Texas project north of Austin; additional satellite parcels have been disclosed in Belton. Less ERCOT congestion than the I-35 South corridor.
Caldwell & Lee Counties
Spillover counties. Cheaper land, fewer disclosure expectations, and proximity to Bastrop have produced quiet speculation and a handful of zoning inquiries; nothing publicly announced at hyperscale scale through May 2026.
Grimes County (adjacent)
The supply chain. Outside Central Texas proper, but the proposed SpaceX Terafab semiconductor plant ($55–$119 B) sits in Grimes and is part of the same AI-buildout footprint. Water, transmission, and tax-incentive questions match those Central Texas counties already face.

What the region looks like in aggregate

Across these counties the Austin Business Journal puts more than 30 publicly identified projects with a combined $50 billion+ in capital investment — almost certainly an undercount, since hyperscale buyers routinely use shell entities and label early acquisitions as “due diligence” until permits force disclosure. The pattern is consistent: power- and water-rich corridors in the I-35 / Highway 130 footprint draw most of the pressure; the Edwards Aquifer recharge zone draws the most contested projects; counties without zoning authority absorb the heaviest siting decisions; and a supply chain — chip fabs, solar-cell factories, transmission — is being assembled around the same footprint.

Sources   Austin Business Journal data-center coverage (2024–26); ERCOT large-load interconnection queue; Bastrop, Travis, Williamson, Bell, Hays, Comal, Caldwell, and Lee county commissioners’ court agendas; PUC SB 1929 registry; GEAA April 2026 report.
LIV Originally May 11 · Updated post-runoff Next dance: November petitions

Data Center Dance — the same old Hustle, or Pay It Forward?

The Hustle, a dance for Boomers, needs a big update. With the May 18 major-party runoffs now behind us, LIV is moving straight to the next dance: helping independents petition onto the November ballot. Susanna Ledesma-Woody (Travis County Commissioner, Pct. 4) carried LIV’s primary backing as a longtime Del Valle volunteer; now the all-volunteer group is pivoting to petition season — with the data-center fight running alongside.

NowPetition support for independent candidates — SE Travis County & Del Valle in focus
OngoingPressure on the data-center buildout — water, energy, transparency, and tax-incentive reform
BriefRachel Hanes’ 62-page GEAA report is the working document — full summary above
“In 2023, the Texas Comptroller estimated this exemption would cost $130 million in FY 2025. Instead, it cost $1.016 billion — one of the most expensive subsidy programs for any industry in any state. Data-center subsidies are now predicted to cost Texas at least $9 billion 2025–2030, and at least $1.3–$1.75 billion annually. This projection is likely to be an underestimate.” GEAA Report, p. 23 · excerpt highlighted by LIV

LIV is all-volunteer; dues start at $10/year. “The door is open for all those weary of the same old hustle who prefer to pay it forward.” — Reprinted from LIV’s May 11 message, livtx.org.

Methodology: This page condenses a 62-page policy report published by the Greater Edwards Aquifer Alliance (GEAA), a nonprofit serving 21 Central Texas counties. Statistics, projections, and recommendations are summarized from that report and its cited sources, including ERCOT, the Texas Comptroller, the U.S. Department of Energy, the Texas Water Development Board, and academic and trade publications. The technologies-in-depth, regional-survey, and orbital-compute sections draw additionally on U.S. Department of Energy data-center efficiency guidance, ASHRAE TC 9.9 thermal guidelines, ERCOT large-load interconnection rules, peer-reviewed carbon-aware computing research, and Austin Business Journal coverage of the SpaceX S-1 filing and Central Texas data-center projects. AustinMayor.com is an independent civic-reference site and is not affiliated with GEAA or any of the companies named above. Framing is descriptive; this page takes no position on whether recommendations should be enacted.
Live