This Month's Latest Tech News in San Antonio, TX - Sunday August 31st 2025 Edition

Weekly Commentary: San Antonio at the intersection of AI growth and local resilience - San Antonio is riding the AI infrastructure wave: projects like Rowan Digital's new Cinco campus (a 300 MW, 440‑acre build near Lytle) are pushing investment and jobs, while coverage from the Texas Matters podcast episode on AI and Texas water and other reporting sound a cautionary note about resources - ERCOT forecasts a need to more than double capacity by 2031 and statewide data‑center water use is already staggering (Newsweek estimates 49 billion gallons in 2025, with local facilities consuming 463 million gallons in 2023–24).

The policy and engineering fix is multi‑pronged - better siting, recycled water, grid planning - and workforce readiness matters: practical courses like the Nucamp AI Essentials for Work bootcamp teach teams how to apply AI responsibly and boost productivity so communities can capture economic benefits without draining wells.

Project	Capacity	Site Size	Investment	Jobs (Const./Perm.)	Target Online
Cinco data center (Rowan)	300 MW	~440 acres	Minimum $900M	~600 / 40+	2027

“Energy demand gets most of the attention, but water is the hidden constraint.”

UTSA's deep dive into “package hallucinations” shines a bright - and worrying - spotlight on how trusted coding copilots can invent entire libraries and hand developers an attack vector on a silver platter: across more than 2.2 million generated code samples the team found 440,445 references to non‑existent packages and 205,474 unique phantom names, with commercial GPT‑series models hallucinating at about 5.2% versus ~21.7% for open‑source models; Python was slightly safer than JavaScript but both were exposed.

The paper, summarized by UTSA on UTSA package hallucinations summary on ScienceDaily and discussed in a technical writeup for the USENIX Security Symposium paper “We Have Package You”, shows these fake recommendations aren't random flukes but often repeat - creating a reliable lure for what reporters now call “slopsquatting,” where attackers publish malicious packages that match the hallucinated names.

The upshot: a casual “pip install” prompted by an AI can become a supply‑chain compromise unless tooling and review practices tighten immediately.

Metric	Value
Referenced hallucinated packages	440,445 of 2.23M samples
Unique invented package names	205,474
Hallucination rate (GPT‑series)	5.2%
Hallucination rate (open‑source)	21.7%
Persistence (regenerated all 10×)	~43%

“It doesn't take a convoluted set of circumstances or some obscure thing to happen,” Spracklen said.

Follow-up reporting and developer action on slopsquatting have to move in lockstep: watchdogs and local reporters must chase the leads that AI tools can invent, while engineers build safer defaults that make phantom packages harder to exploit.

The scale of the reporting challenge is stark - Poynter's summary of the Muck Rack/Rebuild Local News study shows a dramatic shrinkage in local reporting capacity (a 75% drop in journalists per 100,000 since 2002 and more than a third of counties with fewer than the equivalent of one full‑time reporter), so targeted funding and readership support are urgent.

Practical support ideas are already being floated - from the concrete donor and subscription advice in Matt Pearce's guide to backing investigative work to newsroom subsidies - and Cision's State of the Media findings (53% of journalists now use generative AI for tasks like research and transcription) underscore that reporters will use AI, but need better tools and clearer provenance.

That means developers should prioritize verifiable hyperlinks, package‑name checks, and signals that force human review before “pip install” becomes a one‑click risk, while the public and funders shore up the reporting capacity to follow the trail.

“This new data confirms that the local journalist shortage is more severe and far‑reaching than we feared.”

Texas lawmakers advance HB 149 to regulate AI use and government deployment - signed into law on June 22, 2025, the Texas Responsible Artificial Intelligence Governance Act (TRAIGA, HB 149) creates a broad, state‑level rulebook that kicks in January 1, 2026 and covers anyone who develops, deploys or offers AI systems in Texas.

The statute uses a wide definition of “AI system,” tightens biometric consent rules, bars government social‑scoring and AI intended to unlawfully discriminate or incite harm, and requires clear disclosure when state agencies use AI; enforcement is vested solely with the Texas Attorney General (there's no private right of action).

The Act also establishes a 36‑month regulatory sandbox with quarterly reporting and safe harbors for red‑teaming and NIST‑aligned practices, while civil penalties can reach up to $200,000 per uncured violation and daily fines for continued breaches.

Companies and public agencies facing this short runway should inventory AI uses, document purposes and mitigations, and consider the sandbox; for a detailed summary see the InsideTechLaw write‑up on the Texas Responsible Artificial Intelligence Governance Act and DLA Piper's overview of the sandbox and enforcement framework.

Waste Management opens $72M New Braunfels recycling facility using AI optical sorting - The Mesquite Creek Recycling Facility in New Braunfels is a high‑tech material recovery hub built to serve a rapidly growing region that previously had limited recycling access; the $72 million site can process up to 144,000 tons a year and stitches together conveyors, AI‑driven optical sorters and technician oversight so machines spot and divert materials with camera-and-laser precision.

The plant - part of WM's broader multi‑billion dollar investment drive - features more than 16 optical sorters and an extra optical sort line for final quality control, compacts recyclables into bales that can weigh up to 2,000 pounds, and includes an on‑site education center to boost community participation; WM framed the project in its Earth Month update and local coverage shows the facility opened in mid‑April to fanfare from city leaders.

Learn more via the Waste Management official announcement about the Mesquite Creek Recycling Facility, Community Impact San Antonio regional reporting on the New Braunfels recycling facility, and Spectrum / SBG San Antonio coverage of the recycling facility opening.

Item	Detail
Facility	Mesquite Creek Recycling Facility, New Braunfels, TX
Cost	$72 million
Capacity	Up to 144,000 tons/year
Equipment	16+ optical sorters (cameras & lasers); final QC sort line
Address / Opening	801 Kohlenberg Road; grand opening April 15, 2025

“That technology can help us recycle more material, make it cleaner for our customers and really transforms the experience for our employees.”

Meta signs 600MW solar PPA near San Antonio to power data center growth - Tech infrastructure and clean energy intersect with Enbridge's Clear Fork project, a 600 MW solar array under construction near San Antonio for which Meta has agreed to buy 100% of the renewable output under a long‑term PPA; the project carries an estimated US$0.9 billion price tag and is targeting summer 2027 commercial operation, part of a wave of corporate PPAs that are reshaping how data centers secure power.

The move follows Enbridge's final investment decision and positions Clear Fork as a major clean‑energy supply for Meta's regional data center expansion; read the company announcement on Enbridge and coverage from PV‑Tech for project context and timing.

Item	Detail
Project	Clear Fork Solar
Capacity	600 MW
Offtaker	Meta Platforms, Inc. (100% of output)
Estimated cost	US$0.9 billion
Location	Near San Antonio, Texas
Expected in‑service	Summer 2027
Construction status	Underway

“We are thrilled to partner with Enbridge to bring new renewable energy to Texas and help support our operations with 100% clean energy,” - Urvi Parekh, Head of Global Energy, Meta.

AI boom's hidden cost: increased water use in Texas data centers - The rush to build GPU farms is reshaping San Antonio's resource map: two local centers used 463 million gallons of water in 2023–24 and statewide use is already projected at about 49 billion gallons in 2025, ballooning toward roughly 399–400 billion by 2030 as AI workloads scale, according to reporting and a Center for Media Engagement study on data center water use that tracks public attitudes and resource projections; those figures matter because a typical mid‑size facility can sip ~300,000 gallons a day (large sites may need millions), roughly the daily water use of thousands of households.

Communities and utilities face a real tradeoff between jobs and taps, and transparency is thin - Newsweek reporting on Texas data center water use and local reporting show few legal limits on data center water draws - so planning for closed‑loop cooling, reclaimed water and clearer disclosure will determine whether growth becomes a lifeline or a drain for the region.

Read the study summary at the Center for Media Engagement and Newsweek's reporting on Texas water use for context.

Metric	Value
San Antonio data centers (2023–24)	463 million gallons
Texas projected (2025)	~49 billion gallons
Texas projected (2030)	~399–400 billion gallons
Mid‑size data center daily use	~300,000 gallons/day
Large data center daily use (upper range)	~4,500,000 gallons/day

“That's a lot of water, and quite frankly, it's a bit alarming because we are already a state struggling with our water supplies.”

St. Mary's University wins DoD grant to build AI defenses for cyber‑physical systems - a timely award that plugs a local research effort into a national push to harden the “muscles and nerves” of autonomous systems: sensors, actuators and control loops that power drones, robots and critical infrastructure.

The project arrives as the Department of Defense doubles down on AI and autonomy - framing the technology as a transformational force in future warfighting - and follows high‑profile moves like the Pentagon's multibillion‑dollar investments and recent awards to commercial AI providers; the work at St.

Mary's will focus on defensive algorithms, red‑teaming and verification tools meant to prevent adversarial exploits and supply‑chain tricks that could turn benign code into a weapon.

For San Antonio's tech ecosystem, the grant creates an applied pathway from university labs to fielded resilience, aligning local talent with DoD priorities and the broader industry shift toward unmanned, AI‑enabled systems.

DoD AI Funding Highlights	Detail
Pentagon awards for commercial AI tools	Pentagon awards up to $200M in AI contracts to Anthropic, Google, OpenAI, and xAI - NextGov coverage
RDT&E contract example	Metron Inc. $48,652,798 RDT&E award (with $3,000,000 obligated) - U.S. Department of Defense contract notice

“The United States must continue to lead the way in the responsible and ethical development of artificial intelligence – and more specifically, … the Department of Defense must remain at the forefront of the military application of this game‑changing technology.”

Wytec pilots AI gunshot and drug sensing system with 30-day trials for TXShare members - San Antonio–based Wytec is rolling out a phased pilot that stitches distributed AI, smart sensors and a secure private LTE/CBRS backbone into a single Integrated Public Safety Solution, offering TXShare members a 30‑day no‑cost trial to test gunshot and drug sensing in real school settings; the program leverages partnerships with Nextivity and Lemko, more than 75 global patents, and an MSA tied to the NCTCOG/TXShare cooperative that can reach hundreds of jurisdictions (over 170 cities, 50 counties and more than 1,200 ISDs).

The company says the platform feeds location, video and audio to administrators via a mobile app and - based on lab work and field testing - claims better‑than‑90% gunshot identification across 500,000 samples while avoiding automatic 911 calls so teachers and staff remain in the decision loop.

For background on the pilot and TXShare rollout see Wytec's program summary and sales kickoff announcement, and a technical overview of the private‑network approach driving the deployments.

Item	Detail
Trial	30‑day no‑cost trial for TXShare members
Reach	170+ cities, 50 counties, ~1,203 ISDs (TXShare footprint)
Partners	Nextivity Inc., Lemko Corporation
Patents	75+ global patents / patent‑pending tech
Reported accuracy	>90% across 500,000 test samples

“If you're very familiar with some of our competitive technologies, they get too many false positives,” Sanchez said.

UT Health San Antonio develops AI‑guided tai chi app to prevent falls in dementia patients - the concept maps directly onto growing evidence that markerless motion capture (MMC) and smartphone pose estimation can objectively track gait and balance in neurodegenerative conditions, making low‑friction home interventions possible; a JMIR Aging systematic review highlights MMC's promise for full‑body tracking and gait analysis in dementia and mild cognitive impairment (JMIR Aging systematic review of markerless motion capture for dementia gait analysis (2024)), while a Journal of NeuroEngineering and Rehabilitation review shows MMC's clinical potential and the increasing use of smartphones and Kinect‑style cameras for movement assessment (Journal of NeuroEngineering and Rehabilitation review of markerless motion capture applications (2023)).

An AI‑guided tai chi app that uses standard cameras to spot subtle postural drift and step‑timing changes could flag rising fall risk long before a visible tumble, turning everyday practice into continuous, objective monitoring that supports clinicians and caregivers without lab equipment.

Finding	Detail
JMIR Aging review	26 eligible MMC studies; gait analysis in 24 studies (JMIR Aging, 2024)
JNER review	65 MMC studies; Kinect most common, smartphone/video used in 12 studies (JNER, 2023)
MMC modalities	Depth cameras, standard video, mobile phone cameras with deep‑learning postprocessing

Bexar County invests $20M in next‑gen flood warning system with AI‑enabled prediction - County leaders unanimously approved roughly $20 million to build the NextGen Flood Warning System in partnership with the City of San Antonio and the San Antonio River Authority, a two‑year effort that layers upgraded rain gauges, stream sensors, radar and predictive modeling so officials can close low‑water crossings and warn first responders before water overwhelms roadways; the move responds to the June Beitel Creek disaster that swept cars off Loop 410 after heavy, fast‑moving rain and left 13 people dead, and planners say the system will feed real‑time alerts to Bexar Flood official alerts and navigation apps like Waze navigation app while automating gates and lights at the most dangerous crossings.

Officials emphasize rapid data, sensor maintenance (about 30 of 200 gauges need reactivation) and regional modeling to turn responders into pre‑responders, and reporting from Texas Public Radio report and the KSAT news report outlines the technical scope and the push to secure remaining funding from city and state partners.

Item	Detail
Investment	$20 million
Lead agencies	Bexar County, City of San Antonio, San Antonio River Authority
Timeline	About 2 years
Key upgrades	Enhanced gauges & sensors, radar, predictive modeling, automated gates, app/website alerts
Operational goal	Earlier warnings and automated road closures to reduce low‑water crossing fatalities

“AI seems to be in every single conversation that we have, regardless of the topic.”

Conclusion: What this week means for San Antonio's tech future - Momentum is arriving with a clear checklist: funding, skills, and practical security. State and federal programs like the State and Local Cybersecurity Grant Program (SLCGP) - which directs roughly $40M to Texas across 2022–2025 and requires increasing matching funds and an 80% pass‑through to local governments - create a tangible funding stream for city halls, school districts and water utilities to harden systems and train staff; details are available from the Texas SLCGP overview at the Department of Information Resources (Texas DIR overview of the State and Local Cybersecurity Grant Program (SLCGP)).

Local capacity building events like the San Antonio API Security Summit 2025 event information show practitioners sharing hard lessons on real‑world threats, and workforce initiatives - from Impact SA grant cycles to university NSF projects - mean more training pipelines are starting to line up.

Practical, short‑term upskilling matters now: programs such as Nucamp's Nucamp AI Essentials for Work bootcamp (AI at Work: Foundations, Writing AI Prompts, Job Based Practical AI Skills) turn policy dollars and summit conversations into staff who can apply AI responsibly, audit supply chains and keep services running - so the city's next tech win won't just be a new data center, but safer, better‑trained teams ready to manage it.

SLCGP Fact	Detail
Texas allocation (2022–2025)	Approximately $40 million
Matching funds	Required and increasing (e.g., 10% FY22; 20% FY23)
Pass‑through to local governments	Minimum 80% of allocations
Workforce focus	Objective 4 - Workforce Development projects supported