Texas Inference Opens Founding Reservations for Encrypted AI Compute Service

AUSTIN, TEXAS, October 10th, 2026, FinanceWire

The founder and CEO of Texas Inference argues that the $3.7 trillion in annual AI revenue the buildout requires will come mainly from companies reallocating budgets they already have.

Texas Inference is now taking founding reservations for its encrypted AI compute service, which keeps customer data encrypted while it runs and is stored in Texas. The company offers dedicated lanes running a frontier open-source model for teams, as well as encrypted NVIDIA B300 GPUs by the day, week, month or quarter.

Cooper Scanlon, founder and CEO of Texas Inference, is weighing in on one of the most debated questions in artificial intelligence: who will ultimately pay for the trillions of dollars being spent to build it. His answer is that the bill will fall mainly on businesses rather than households.

Scanlon’s analysis responds to a new paper, “Financing the AI Buildout,” by Columbia Business School professor Stijn Van Nieuwerburgh, presented at the Brookings Papers on Economic Activity conference in September.

The $3.7 Trillion Question

The paper estimates that US investment in AI infrastructure will reach about $10.3 trillion between 2025 and 2032. To earn a 10% return on that investment, AI revenue would need to reach roughly $3.7 trillion a year by 2032, or about 9.2% of projected GDP. By comparison, OpenAI and Anthropic together currently earn around $100 billion a year.

The figure has prompted wide discussion online, with some commentators framing it as an amount Americans will eventually need to spend on AI services each year. Scanlon sees the money coming from a different place.

Why Households Cannot Carry It

Scanlon notes that a household version of the 9.2% figure would amount to about $11,000 per American every year, a share of GDP comparable to what the country spends on food. “US consumers simply do not have this much extra money to spend,” he wrote.

Instead, he points to businesses, which have driven around three-quarters of computing spending. Citing Bureau of Economic Analysis data, he notes that business investment in computers and peripherals reached an annualized $325 billion in late 2025, up 75% from $186 billion in 2024.

Most Businesses Have Not Started Yet

Much of the spending so far, Scanlon observes, has come from a small group of very large companies. The buildout was first funded from cash held by firms such as Amazon, Meta and Google, and is increasingly moving to outside financing.

Citing estimates from Apollo, he notes that S&P 500 companies employ up to 18% of US workers, account for 21% of US capital expenditure and generate about half of corporate profits, while 81% of firms with more than $100 million in revenue are private. In his view, most of the S&P 500 and most private businesses have not yet bought or integrated AI at scale, which leaves considerable room for growth.

Reallocating Existing Budgets

Beyond the new value AI creates, Scanlon expects much of the money to come from budgets companies already control. He notes that US employers paid about $16 trillion in wages and benefits at the end of 2025, and that workers’ compensation has held near 52% of gross domestic income since 2022. If that trend holds, a $40 trillion economy in 2032 would see wages and benefits of around $21 trillion, making $3.7 trillion roughly 18% of that pool.

Corporate profits have reached a record 12.1% of GDP, he adds, but employee budgets remain the larger pool, and one that companies already know how to reallocate.

Scanlon wrote that the 9% figure is “most likely to be driven by businesses shifting employee spend to AI spend.” He added, “Businesses must invest in AI to stay competitive. That investment will drive efficiency and profits for them by cutting costs elsewhere.”

The Case for Owning Compute

Scanlon’s perspective is shaped by his work at Texas Inference, an encrypted AI compute company that is currently taking founding reservations. Customers can reserve a dedicated lane running a frontier open-source model for teams of 25 to 100 people, or rent between one and eight encrypted NVIDIA B300 GPUs by the day, week, month, or quarter.

The company’s approach is built around keeping data encrypted while it runs. Intel TDX protects the memory of the host machine, Blackwell hardware encrypts the memory of each GPU, and traffic between GPUs is encrypted as well. Customer data stays in Texas, and the hardware signs a report for every session so customers can independently verify those protections.

As AI moves deeper into the core operations of companies, Scanlon has argued that businesses will increasingly need dedicated compute of their own to avoid enterprise software markups and protect sensitive data. Latham and Watkins, the 2nd largest US Law Firm, cited these as driving factors for their own purchasing of in-house GPU servers. Confidential computing, which keeps data encrypted even while it is being processed, sits at the center of that view.

Whether the AI buildout ultimately pays off remains one of the defining economic questions of the decade. For Scanlon, the answer will depend less on consumer subscriptions and more on how quickly businesses move AI from experimentation into the center of how they operate. More about his work is available at texasinference.ai.

About Texas Inference

Texas Inference is an encrypted AI compute company based in Austin, Texas. It provides confidential AI compute in which customer data stays in Texas and remains encrypted while it runs, protected by Intel TDX, Blackwell GPU memory encryption and encrypted traffic between GPUs. The hardware signs a report for every session, allowing customers to independently verify those protections. More information is available at texasinference.ai.

Contact