Scotty Nyquist discuss the growth in AI data centers and the strain on the system. Photo via HARC report

For most of the past 20 years, U.S. electricity policy relied on predictable trends in demand. Electricity use, in most regions, increased gradually, forecasts were stable, and utilities adjusted the system in small steps. Power plants, transmission lines, and substations were generally added to reflect shifts in load, rather than growth, and costs were recovered through modest adjustments to customer bills.

Growth in AI data centers has disrupted this model. A single facility can add as much electricity demand as a small town. That demand comes all at once, runs continuously, and has little tolerance for outages. If electricity service drops even briefly, computation stops, and services shut down. Ironically, data centers need reliable service, a point that their emergence is driving concern around for the rest of the grid.

What the numbers say

The International Energy Agency projects global electricity consumption from data centers to double by 2030, reaching roughly 945 TWh, nearly 3 percent of global electricity demand, with consumption growing about 15 percent per year this decade. McKinsey projects that U.S. data center demand alone could grow 20–25 percent per year, with global capacity demand more than tripling by 2030.

After years of roughly 0.5 percent annual demand growth, many forecasts now place total U.S. electricity demand growth closer to 2–3 percent per year through the mid-2030s, with much higher growth in specific regions. In Texas, some forecasters are saying electricity demand could double over the next five years, a staggering 10 percent per year growth rate. What sounds incremental on paper translates into a major challenge on the ground. Meeting this pace of growth is estimated to require $250–$300 billion per year in grid investment, about double what the system has been absorbing.

Where the system starts to strain

The strain appears first in the interconnection queue. It shows up as long waits, backlogs, and delays for connecting new loads and new generation.

Before new generators or large load customers can be connected, a study is required to assess their impact on the grid, whether it can physically handle the added load, and whether upgrades are required. With AI-driven data centers, utilities face far more connection requests than they can realistically support. In ERCOT, large-load interconnection requests exceed 200 gigawatts, most tied to data centers. That amount exceeds historical norms, and it is several times larger than what can be practically studied or built in the near term.

To be clear, public utility commissions are required to study these requests because they must manage system capabilities to ensure minimal disruption. This means engineers spend time evaluating projects that may never be built, while other more commercially viable projects may wait longer for approvals. This extends timelines and makes infrastructure planning less reliable.

Why policymakers are rethinking the rules

Utilities and their regulators must decide how much generation, transmission, and substation capacity to build years before it comes online. Those decisions are based on expected demand at the time projects are approved. When it comes to data centers, by the time infrastructure is completed, they may end up deploying newer, more efficient chips that use less power than originally assumed. This can result in grid infrastructure built for a higher load than what actually materializes, leaving excess capacity that still must be paid for through system-wide rates.

That’s the central dilemma. If utilities build too little capacity, the system operates with less reserve margin. During periods of grid stress, operators have fewer options, increasing the likelihood of curtailments or outages. However, if utilities build too much, customers may be asked to pay for infrastructure that is not fully used.

In response, policymakers are adjusting the rules. In some regions, regulators are moving toward bring-your-own-power approaches that require large data centers to supply or fund part of the capacity needed to serve them or reduce demand during system stress. At the federal level, permitting reforms tied to datacenter infrastructure increasingly treat electricity as a strategic economic input.

As Ken Medlock, senior director at the Baker Institute Center for Energy Studies (CES), explains:

“Many of the planned data centers are now also adding behind-the-meter options to their development plans because they do not anticipate being able to manage their needs solely from the grid, and they certainly cannot do so with only intermittent power sources.”

Behind-the-meter (BTM) refers to power that a consumer controls on its side of the utility meter, such as on-site gas generation or a dedicated power plant. These resources allow data centers to keep operating during grid-related service. Most facilities remain connected to the grid, but the backup BTM generation serves as insurance for operating their core business.

This shifts responsibility. Utilities traditionally manage reliability across all customers by maintaining an operating reserve margin, or spare capacity. Increasingly, large-load customers manage part of their own electricity reliability needs, which changes how infrastructure is planned and how risk is distributed.

Bottom line

AI-driven load growth is arriving faster and in more concentrated places than the power system was built to accommodate. Utilities and regulators are being forced to make decisions sooner than planned about where to build, how fast to build, and which customers get priority when capacity is limited. The effects extend beyond data centers, showing up in system costs, reliability margins, competition for grid access, and pressure on communities and industries that depend on affordable and dependable power. The issue is not whether electricity can be generated, but how the costs and risks of rapid demand growth are distributed as the system tries to keep up. How regulators balance these decisions will determine who pays as AI demand outruns the power grid.

-----------

Scott Nyquist is a senior advisor at McKinsey & Company and vice chairman, Houston Energy Transition Initiative of the Greater Houston Partnership. The views expressed herein are Nyquist's own and not those of McKinsey & Company or of the Greater Houston Partnership. This article originally appeared on LinkedIn.

Collide has rolled out RIGGS, a large language model for energy professionals. Photo via Getty Images

Houston AI startup rolls out platform to reshape oil and gas workflows

AI for energy

Houston-based Collide is looking to solve AI issues in the energy industry from within.

Co-founded by former oil roughneck Collin McLelland, the company has developed AI software for operators and field teams, shaped by firsthand oilfield experience. Its AI-native platform “retrieves and synthesizes data from authoritative sources to deliver accurate, cited, and energy-focused insights to oil and gas professionals,” according to the company.

“Oil and gas has a graveyard full of technology that was technically impressive and operationally useless,” McLelland tells Energy Capital. “The reason is almost always the same: the people who built it didn't understand what they were actually solving for. When you're an outsider, you see workflows and try to automate them. When you're an insider, you understand why those workflows exist—the regulatory constraints, the physical realities, the liability concerns, the trust dynamics between operators and service companies.”

Collide’s large language model, known as RIGGS, performed well in recent benchmarking results when taking a standardized petroleum engineering (SPE) exam, the company reports. The exam assesses understanding from conceptual terminology to complex mathematical problem-solving.

According to Collide, RIGGS achieved a score of 67.5 percent on a 40-question subset of the SPE petroleum engineering exam, outperforming other large language models like Grok 4 (62.5 percent), Claude Sonnet 4.5 (52.5 percent) and GPT 5.1 (4 percent).

RIGGS completed the test in 15 minutes, while Grok took two hours. Collide hopes over the next few months, RIGGS will receive a score between 75 percent to 80 percent accuracy.

The software could potentially help oil and gas companies produce accurate outputs and automate trivial workflows, which can open up valuable time for engineers and teams to work on other pressing matters, according to McLelland.

“Collide exists because we sat in those seats — we were the engineers, the operators, the field guys,” he says. ”RIGGS scoring higher on the PE exam versus the frontier labs isn't a party trick. It's evidence that the model understands petroleum engineering the way a petroleum engineer does, because it was built by people who do.”

RIGGS was trained on Collide’s Spindletop hardware and is supported by a vast library of information, as well as a reasoning engine and validation layer that uses logic to solve problems.

“Longer term, we see RIGGS as the intelligence layer that sits underneath every operator's workflow — not a chatbot you open in a browser, but something embedded in the tools engineers already use,” McLelland says. “The goal is to give every engineer the knowledge and pattern recognition of a 30-year veteran, on demand."

According to McLelland, Collide is already building toward reservoir analysis and production optimization, automated regulatory compliance (Railroad Commission filings, W-10s, G-10s), workover report generation, and engineering decision support in the field for near-term use cases. In March, Collide and Texas-based oil and gas operator Winn Resources announced a collaboration to automate the time-intensive process of filing monthly W-10 and G-10 forms with the Texas Railroad Commission, completing what’s normally a multi-hour task in under 30 minutes. Collide reports that Winn’s infrastructure now automates regulatory filings and provides real-time visibility into data gaps, which has reduced processing time by over 95 percent.

“Before Collide, I'd spend hours manually keying in filings,” Buck Crum, director of operations, said in a news release. “(In March), we had 50 wells to file and I was done in 20 minutes. It does the majority of the heavy lifting while keeping me in control. That human-in-the-loop approach saves meaningful time and gives us greater confidence in our compliance and reporting.”

Collide was originally launched by Houston media organization Digital Wildcatters as “a professional network and digital community for technical discussions and knowledge sharing.” After raising $5 million in seed funding led by Houston’s Mercury Fund last year, the company said it would shift its focus to rolling out its enterprise-level, AI-enabled solution.
Merab Momen, founder of AI CTO Services. Courtesy Photo

How this Houston expert helps startups turn AI hype into real impact

now streaming

Artificial intelligence is now everywhere. It is mentioned in every startup pitch deck, and every corporate roadmap claims to use it. However, many early-stage businesses struggle with the simple question, “What does AI actually mean for my business?”

In a recent podcast episode of EnergyTech Startups, Merab Momen, founder of AI CTO Services and a long time AI practitioner, explains why most founders misunderstand AI, how startups can practically apply it and why Houston is quietly becoming a serious hub for AI-driven innovation.

Filling the AI Leadership Gap

Merab’s career has spanned decades of technology transitions. He worked on neutral networks in the 1990s, constructed computer vision systems long before they were common, and helped install AI solutions inside huge industrial companies. However, he noticed a huge problem when generative AI started to explode into the mainstream-The requirement of a real partner by the founders for AI integration but inability to rely on a full-time CTO and project-based consultants.

“I really needed something which is much more engaging where I can give that partner-level advice to the founders,” he said. By giving firms on-demand access to high-level AI knowledge and expertise, his methodology enables them to analyse tools, steer clear of cost blunders and eventually transition to a permanent technology leader when the time is right.

AI is Older than Most People Think

Despite its recent rise in popularity, AI is nothing new. AI actually began in the 1950s. Merab in his conversation explained how he worked on his first AI project back in the year 1996 that worked perfectly, but the processing power wasn’t just there to make it practical. He continued how he utilized the swarm intelligence models to optimize supply chains, now referred to as MLPOs and data engineering.

From Language Models to Physical World

Much of the public conversation about AI revolves around chatbots and text generation. But Merab sees far greater potential in AI’s interaction with the physical world, especially in industrial settings. He emphasized edge computing and vision language models (VLMs) as significant advances in manufacturing and energy. This physical shift is opening doors for new opportunities for robotics, automated inspections, and industrial safety applications. Merab added that Houston is uniquely positioned for this transition.

Why Houston has an AI Advantage

Silicon Valley may dominate the AI headlines, but Merab believes Houston’s advantage lies beneath the surface. The city doesn’t lag in AI utilization; it just operates in industries where results show differently.

Machine learning isn’t new to Houston’s core industries. Energy companies, manufacturers, logistics providers, and healthcare systems have been using advanced analytics for decades. The difference lies in them innovating in industrial sectors rather than consumer technology.

What’s Next

With the AI CTO Services growing, Merab is working with startups across industries to deploy AI in practical, business-first ways.

He is more interested in assisting founders in finding answers to critical issues than following new trends.

For Houston’s energy and climate tech community, it needs to transform AI enthusiasm into real-world impact.

Listen to the full conversation with Mehrab Momin on the Energy Tech Startups Podcast to learn more.

---

Energy Tech Startups Podcast is hosted by Jason Ethier and Nada Ahmed. It delves into Houston's pivotal role in the energy transition, spotlighting entrepreneurs and industry leaders shaping a low-carbon future.


A new report shows the role Texas could play as the data-center sector enters "hyperdrive." Photo via JLL.com.

Texas could topple Virginia as biggest data-center market by 2030, JLL report says

data analysis

Everything’s bigger in Texas, they say—and that phrase now applies to the state’s growing data-center presence.

A new report from commercial real estate services provider JLL says Texas could overtake Northern Virginia as the world’s largest data-center market by 2030. Northern Virginia is a longtime holder of that title.

What’s driving Texas’ increasingly larger role in the data-center market? The key factor is artificial intelligence.

Companies like Google and Microsoft need more energy-hungry data centers to power AI innovations. In a 2023 article, Forbes explained that AI models consume a lot of energy because of the massive amount of data used to train them, as well as the complexity of those models and the rising volume of tasks assigned to AI.

“The data-center sector has officially entered hyperdrive,” Andy Cvengros, executive managing director at JLL and co-leader of its U.S. data-center business, said in the report. “Record-low vacancy sustained over two consecutive years provides compelling evidence against bubble concerns, especially when nearly all our massive construction pipeline is already pre-committed by investment-grade tenants.”

Dallas-Fort Worth has long dominated the Texas data-center market. But in recent years, West Texas has emerged as a popular territory for building data-center campuses, thanks in large part to an abundance of land and energy. Nearly two-thirds of data-center construction underway now is happening in “frontier markets” like West Texas, Ohio, Tennessee and Wisconsin, the JLL report says.

Northern Virginia, the current data-center champ in the U.S., boasted a data-center market with 6,315 megawatts of capacity at the end of 2025, the report says. That compares with 2,423 megawatts in Dallas-Fort Worth, 1,700 megawatts in the Austin-San Antonio corridor, 200 megawatts in West Texas, and 164 megawatts in Houston.

UH researchers have developed a thin film that could allow AI chips to run cooler and faster. Photo courtesy University of Houston.

Houston researchers develop energy-efficient film for AI chips

AI research

A team of researchers at the University of Houston has developed an innovative thin-film material that they believe will make AI devices faster and more energy efficient.

AI data centers consume massive amounts of electricity and use large cooling systems to operate, adding a strain on overall energy consumption.

“AI has made our energy needs explode,” Alamgir Karim, Dow Chair and Welch Foundation Professor at the William A. Brookshire Department of Chemical and Biomolecular Engineering at UH, explained in a news release. “Many AI data centers employ vast cooling systems that consume large amounts of electricity to keep the thousands of servers with integrated circuit chips running optimally at low temperatures to maintain high data processing speed, have shorter response time and extend chip lifetime.”

In a report recently published in ACS Nano, Karim and a team of researchers introduced a specialized two-dimensional thin film dielectric, or electric insulator. The film, which does not store electricity, could be used to replace traditional, heat-generating components in integrated circuit chips, which are essential hardware powering AI.

The thinner film material aims to reduce the significant energy cost and heat produced by the high-performance computing necessary for AI.

Karim and his former doctoral student, Maninderjeet Singh, used Nobel prize-winning organic framework materials to develop the film. Singh, now a postdoctoral researcher at Columbia University, developed the materials during his doctoral training at UH, along with Devin Shaffer, a UH professor of civil engineering, and doctoral student Erin Schroeder.

Their study shows that dielectrics with high permittivity (high-k) store more electrical energy and dissipate more energy as heat than those with low-k materials. Karim focused on low-k materials made from light elements, like carbon, that would allow chips to run cooler and faster.

The team then created new materials with carbon and other light elements, forming covalently bonded sheetlike films with highly porous crystalline structures using a process known as synthetic interfacial polymerization. Then they studied their electronic properties and applications in devices.

According to the report, the film was suitable for high-voltage, high-power devices while maintaining thermal stability at elevated operating temperatures.

“These next-generation materials are expected to boost the performance of AI and conventional electronics devices significantly,” Singh added in the release.

---

This article originally appeared on our sister site, InnovationMap.

Ad Placement 300x100
Ad Placement 300x600

CultureMap Emails are Awesome

Texas City ammonia plant acquired by Yara in $1.3 billion deal

Ammonia Acquisition

Yara North America, a subsidiary of Norwegian fertilizer and ammonia producer Yara International, has agreed to buy an ammonia production plant in Texas City for $1.3 billion.

The seller is GCA Holdings, an affiliate of Texas City-based chemical manufacturer Gulf Coast Ammonia, which is owned by private equity firms Lotus Infrastructure Partners and MB Energy.

The Texas City plant, with an eventual annual capacity of 1.3 million metric tons, is expected to start full production by the end of this year. Yara says the ammonia produced by the plant will serve its own fertilizer production system and its key customers.

During a recent call with analysts and investors, Magnus Ankarstrand, executive vice president and CFO of Yara International, said the plant holds the potential to become one of the company’s most profitable plants. The $1.3 billion purchase price, he added, “is a very attractive entry ticket to ammonia production in the U.S. at a very attractive cost.”

The Texas City plant will add to Yara’s holdings in the Lone Star State, as Yara is the majority owner of an ammonia, hydrogen and nitrogen production plant in Freeport.

Construction of the ammonia plant began in 2020, but technical and infrastructure issues delayed the project. On its website, Gulf Coast Ammonia says the plant represented a $600 million investment.

“Gulf Coast Ammonia is a world-class asset that required disciplined execution across development, financing, construction, and commercial structuring,” Philipp Pletka, managing director of Lotus Infrastructure Partners, says in a news release.

Trexlertown, Pennsylvania-based Air Products, which owns and operates the country’s largest hydrogen pipeline network, will continue to supply hydrogen and nitrogen for the plant under a long-term deal with Yara, according to the release.

However, the news comes two days after Yara International announced that it would no longer be purchasing ammonia assets in the Louisiana Clean Energy Complex (LCEC) from Air Products. In a separate release, Yara said it planned to reallocate funds toward "alternative mature U.S. ammonia investment opportunities with more competitive returns."

Houston hypersonic engine company lands $91M to accelerate production

Clean Speed

Houston-based Venus Aerospace has closed a $91 million Series B round and plans to scale the production of its hypersonic engine.

The round was led by Houston-based Mercury Fund with participation from Lockheed Martin Ventures, MESH, PEAK6, Draper Associates, Starboard Star Venture Capital, Green Sands Equity and other investors, according to a news release.

The investment comes about a year after Venus completed the first U.S. flight test of its high-thrust rotating detonation rocket engine (RDRE). The engine is expected to enable vehicles to travel four to six times the speed of sound from a conventional runway and is about 15 percent more efficient than traditional alternatives, according to the company.

Venus Aerospace says the latest round of funding will allow it to move the RDRE from demonstration to deployment and meet customer requirements for the near-term defense and space industries. The company says that the reusable RDRE is designed with a "common propulsion architecture" that can work for multiple industries and mission types.

“This financing marks an important step in moving Venus from breakthrough demonstration to scaled capability,” Sassie Duggleby, co-founder and CEO, said in the news release. “Our customers need propulsion systems that go farther, can be produced reliably and are built on supply chains they can trust. We are advancing that capability with American engineering and manufacturing talent to strengthen U.S. defense, expand space access and support the future of high-speed flight.”

Venus Aerospace raised a $20 million Series A in 2022, led by Wyoming-based Prime Movers Lab. At the time, the company said it would put the funding toward three main technologies: a next-generation rocket engine, aircraft shape and leading-edge cooling system.

The company also picked up an investment from Lockheed Martin Ventures, the investment arm of aerospace and defense contractor Lockheed Martin, in November 2025—in addition to funding from other investors over the years.

“Since our initial investment, Venus has progressed very quickly in its technology development," Chris Moran, vice president and general manager of Lockheed Martin Ventures, added in the release. "Our reinvestment in Venus recognizes Venus’ accomplishments to date and focus on speed to manufacture, cost management and reduction of supply chain constraints. Venus is working effectively to position its propulsion system for the production scale required by defense programs.”

"Venus is exactly the kind of company Houston capital should be backing," Blair Garrou, co-founder and managing partner at Mercury Fund, added in the release. "It combines multiple frontier technologies, domestic manufacturing and clear commercial and national security relevance. We believe this team is positioned to lead an important new chapter in defense and space, and we are proud to support a company building breakthrough technology here in Texas."

Venus Aerospace and Houston clean tech startup Vaulted Deep were also named to the World Economic Forum's Technology Pioneers community earlier this summer.

---

This article first appeared on InnovationMap.com.

14 climatech startups join Greentown Houston in first half of 2026

green team

Climatech incubator Greentown Labs reports that 14 startups have joined its Houston community so far this year.

The companies are among 30 new startups to have joined Greentown Houston and Greentown Boston in 2026. Four of the companies are headquartered in Houston.

The startups are working on a range of "hydrogen-powered heavy-duty transport to AI-driven grid interconnection," according to Greentown.

The local startups that joined Greentown Houston include:

  • Houston-based Focis AI, which transforms industrial laser scans into structured asset intelligence to automatically identify, classify and map components in refineries and plants
  • Houston-based Iron Lattice, which develops next-generation memory technology for AI and high-performance computing that improves energy efficiency, endurance and scalability while remaining compatible with existing semiconductor manufacturing
  • Houston-based Orbital Arc, which is developing a new ion engine designed to improve the efficiency and scalability of spacecraft propulsion from low Earth orbit to deep space
  • Houston-based Sustain Energy LLC, which delivers cleaner, lower-cost fuel to industrial customers in pipeline-absent, underserved markets, cutting their energy costs and emissions with no infrastructure investment on their end

Other startups from around the world joined the Houston incubator in the same time period, including:

  • Ankara-based AIS Field, which develops robotic, AI-assisted non-destructive inspection systems, including submersible tank and boiler crawlers
  • San Francisco-based Armada AI, which builds rapidly deployable modular and edge data centers that run on local, stranded, or renewable power
  • San Francisco-based Armeta, which turns complex engineering drawings and legacy documentation into structured, usable data
  • Pittsburgh-based Atlas Robotics, which develops a Physical AI platform that powers autonomous material-handling robots and AI-guided forklifts
  • Ghana-based Cocoa Potash, which transforms high-emissions agricultural waste from cocoa, coconut, and palm-nut into organic potash, fertilizer and renewable energy
  • Israel-based Criaterra, which produces low-carbon, cement-free building materials
  • Italy-based ETAK, which manufactures modular reactors that convert solid waste into clean syngas
  • Kenya-based FelixFusion, which uses its Felix platform to model every grid connection point, including capacity, upgrade costs, and constraints
  • San Diego-based Gemini Energy, which builds next-generation fuel cells for data-center power
  • Tokyo-based Hibot, which develops robotic systems for inspecting and maintaining infrastructure in hazardous, hard-to-access environments
  • Austin-based Sheetak, which designs and manufactures thermoelectric coolers, generators, and assemblies for solid-state cooling and energy harvesting
  • The Netherlands-based ToPerform, which makes AI-powered, non-intrusive fouling sensors that monitor pipelines around the clock and predict the optimal cleaning time

Another 16 startups joined Greentown's Boston incubator. See the full list of new members here.

More than 100 startups joined Greentown last year, according to an end-of-year reflection shared by Greentown CEO Georgina Campbell Flatter. Read more about them here.