The most ambitious data center project in the history of computing is hemorrhaging leadership. Again.
Two senior executives steering OpenAI’s Stargate initiative — the joint venture with SoftBank that aims to pour up to $500 billion into AI infrastructure across the United States — have departed the organization in recent weeks, according to a report from The Information. The exits mark the latest in a pattern of high-profile turnover that has dogged OpenAI’s infrastructure ambitions even as the company races to secure its position as the dominant force in artificial intelligence.
The departures include Kyle Khoury, who served as vice president of data center development for the Stargate venture, and Sammy Sidhu, a senior figure involved in technical infrastructure planning. Both played central roles in translating OpenAI’s enormous capital commitments into physical reality — the actual buildings, power contracts, and cooling systems required to train and run the next generation of AI models. Their exits were not publicly announced.
This matters far more than typical executive churn.
Stargate isn’t a side project. It’s the backbone of OpenAI’s long-term strategy. Announced in January alongside President Trump at the White House, the venture initially committed $100 billion in spending, with aspirations to reach $500 billion over four years. SoftBank’s Masayoshi Son, Oracle’s Larry Ellison, and OpenAI CEO Sam Altman stood together for the announcement, framing it as a national competitiveness imperative — a way to ensure that the infrastructure powering AI’s future would sit on American soil, not in the hands of foreign rivals.
But building that infrastructure requires people who know how to do it. And OpenAI keeps losing them.
The departures of Khoury and Sidhu follow a string of earlier exits from OpenAI’s infrastructure ranks. The company has struggled to retain leaders in a market where experienced data center executives are among the most sought-after professionals in technology. Hyperscalers like Microsoft, Google, and Amazon have spent decades cultivating internal expertise in large-scale facility development. OpenAI, despite its soaring valuation — now reportedly around $300 billion following its recent funding round — is still relatively young as an organization and has faced persistent questions about whether it can execute on physical infrastructure at the scale it has promised.
The timing is particularly awkward. OpenAI is in the middle of converting from a nonprofit-controlled structure to a for-profit public benefit corporation, a transition that has itself generated controversy and legal challenges. The company closed a $40 billion funding round led by SoftBank in April, as reported by The Wall Street Journal, valuing it at $300 billion. That capital is supposed to fuel, among other things, the Stargate buildout. Losing the people responsible for spending that money wisely is not a minor inconvenience.
Data center construction is a brutally complex discipline. It sits at the intersection of real estate, electrical engineering, supply chain logistics, and regulatory compliance. A single large-scale AI training facility can require hundreds of megawatts of power — enough to light a small city. Securing that power, negotiating with utilities, obtaining permits, managing construction timelines, and ensuring the facility meets the thermal demands of thousands of GPUs running simultaneously is work that takes years of specialized experience. It’s not something you learn from a textbook or pick up in a few months on the job.
And the competition for this talent is ferocious. Microsoft alone has committed over $80 billion in capital expenditure for fiscal year 2025, much of it directed at data center expansion to support its Azure cloud and AI workloads. Amazon Web Services, Google Cloud, and Meta are all engaged in their own massive buildouts. According to reporting from Reuters, OpenAI and SoftBank have been actively seeking additional funding sources for Stargate, an indication that even the initial $100 billion commitment may not be sufficient for the scope of what’s planned.
The Stargate venture’s structure adds another layer of complexity. It operates as a joint venture, not a wholly owned subsidiary of either OpenAI or SoftBank. Oracle holds an equity stake and is expected to provide cloud infrastructure. MGX, an Abu Dhabi-backed technology investment fund, is also involved. Coordinating strategy across multiple stakeholders with different priorities, risk tolerances, and corporate cultures is inherently difficult. When the people who understand how all the pieces fit together walk out the door, institutional knowledge walks out with them.
Some context on the broader AI infrastructure race helps explain why these departures reverberate. The International Energy Agency estimated earlier this year that global electricity consumption from data centers could more than double by 2026. In the United States, utilities in Virginia, Texas, and other data center hubs are scrambling to meet demand, with some warning of potential grid constraints. The race to secure power has become so intense that some companies are exploring nuclear energy, including small modular reactors, as a long-term solution. OpenAI itself has expressed interest in nuclear-powered data centers, though such facilities remain years from operational reality.
Against this backdrop, execution speed is everything. Every month of delay in breaking ground on a new facility is a month that competitors can use to expand their own capacity. And in AI, capacity translates directly into capability — the ability to train larger models, serve more customers, and generate more revenue. OpenAI’s leadership understands this. Altman has spoken repeatedly about the need for massive infrastructure investment, calling it a prerequisite for the next leaps in AI performance.
But understanding the need and meeting it are different things.
OpenAI’s organizational turbulence extends well beyond its infrastructure team. The company has seen the departure of co-founders Ilya Sutskever and John Schulman, former CTO Mira Murati, and numerous other senior figures over the past 18 months. The pattern has raised questions among investors and industry observers about whether OpenAI’s internal culture — shaped by rapid growth, intense pressure, and Altman’s demanding leadership style — can sustain the kind of operational continuity that a half-trillion-dollar infrastructure program demands.
Not everyone sees the departures as catastrophic. Large organizations routinely experience leadership transitions, and OpenAI has shown an ability to recruit aggressively. The company’s brand, its access to capital, and the sheer excitement surrounding its mission make it an attractive employer for many. Replacing a vice president of data center development is not the same as replacing a chief scientist. The skills are more transferable, and the talent pool, while tight, is not as narrow as it is for frontier AI research.
Still. Two key Stargate leaders leaving in the same stretch raises legitimate questions about what’s happening inside the venture. Are there strategic disagreements about site selection, spending priorities, or technical architecture? Is the joint venture structure creating friction that makes it hard for leaders to operate effectively? Or is this simply the result of a hot labor market, where executives with data center experience can command enormous compensation packages from rivals?
The answer is probably some combination of all three.
SoftBank, for its part, has been expanding its own infrastructure ambitions independently of the Stargate framework. Son has made AI the centerpiece of SoftBank’s investment thesis, committing tens of billions to chip design, data center development, and AI startups. The conglomerate’s vision doesn’t depend entirely on OpenAI, which may create subtle tensions within the partnership. When two organizations with different strategic horizons share a joint venture, alignment can erode quickly — especially when the people who managed that alignment are no longer around.
Oracle’s role adds yet another variable. Larry Ellison has positioned Oracle Cloud Infrastructure as a serious contender in the AI training market, touting large GPU clusters and competitive pricing. But Oracle’s cloud business remains a distant fourth behind AWS, Azure, and Google Cloud in overall market share. Whether Oracle can deliver the scale of infrastructure that Stargate requires — on time and at the performance levels OpenAI needs — is an open question that the departure of key technical leaders makes harder to answer with confidence.
For the broader AI industry, the Stargate leadership exits are a reminder that the bottleneck in artificial intelligence is shifting. For years, the constraint was algorithmic — could researchers design models capable of genuine intelligence? Then it became a compute constraint — were there enough GPUs? Now, increasingly, the constraint is physical. Power. Land. Water for cooling. Construction crews. Permitting. And the human beings who know how to orchestrate all of it.
OpenAI can raise virtually unlimited capital. It has access to the most advanced chips from Nvidia and is reportedly designing its own custom silicon. Its models are among the most capable in the world. None of that matters if it can’t build the facilities to run them.
The next few months will be telling. If OpenAI quickly fills the vacated roles with experienced leaders and maintains its construction timelines, the departures will be a footnote. If the positions remain open, or if additional infrastructure executives leave, it will signal something deeper — a structural problem in how the Stargate venture is organized and led. Investors who just committed $40 billion to OpenAI’s future will be watching closely. So will the White House, which staked political capital on the Stargate announcement as proof that America’s AI future is secure.
Sam Altman has built OpenAI into the most talked-about company in technology. Building data centers, it turns out, requires a different set of skills than building hype. The question now is whether OpenAI can hold onto the people who possess them.
WebProNews is an iEntry Publication