OpenAI Sells Compute Access Like Futures Contracts

OpenAI announced Tuesday a new offering called Guaranteed Capacity — essentially, enterprise customers can now lock in long-term access to compute power for one, two, or three years, with discounts that scale with commitment length.

CEO Sam Altman framed it simply in a post on X: "Customers are increasingly asking us for certainty on capacity."

As models get bigger and more computationally hungry, access to AI infrastructure is becoming its own competitive moat. Companies that can't secure capacity are left behind.

Altman added that OpenAI will hold back enough compute to keep its own products — ChatGPT and coding assistant Codex — running. Beyond that, it's first-come, first-served until the current allocation sells out.

What This Actually Means

According to CNBC, OpenAI has told investors it's targeting roughly $600 billion in total compute spend by 2030.

That's spending, not revenue. The Guaranteed Capacity program is OpenAI's answer to the obvious follow-up question: how do you justify burning that kind of cash?

You pre-sell the capacity. You get customers to commit years in advance. You turn infrastructure into a subscription business.

It's smart. It's also a sign that OpenAI is under real pressure to show Wall Street a coherent business model ahead of what CNBC describes as a "potentially massive IPO as soon as this year."

The company is currently valued at more than $850 billion by private investors. That number needs a revenue story behind it — and Guaranteed Capacity is part of building one.

Dell Partnership: OpenAI Goes On-Premises

Also new: OpenAI and Dell Technologies announced a partnership on May 18, according to OpenAI's own newsroom, to bring Codex to hybrid and on-premises enterprise environments.

This is a significant shift. OpenAI built its empire on cloud-based, API-delivered AI. Now it's pushing into the data centers that large enterprises and government contractors already own.

A massive slice of the enterprise market — regulated industries, defense contractors, financial institutions — cannot send their data to a third-party cloud. On-premises deployment is the only way OpenAI gets into those accounts.

The Dell deal is OpenAI knocking on doors it previously couldn't open.

The Product Blitz Is Real

Since our last coverage, OpenAI's newsroom has been a firehose. Just in the past few weeks, according to OpenAI directly:

• GPT-5.5 launched April 23
• ChatGPT Images 2.0 dropped April 21
• New voice intelligence models hit the API on May 7
• Ads in ChatGPT began testing May 7
• Codex went to AWS on April 28
• A personal finance feature in ChatGPT launched May 15
• The OpenAI Deployment Company officially launched May 11

This is the pace of a company sprinting toward an IPO finish line.

What Mainstream Coverage Is Getting Wrong

Bloomberg teased a story about SoftBank founder Masayoshi Son's "starstruck bet" on OpenAI raising concerns — but their paywall kept the details locked. The headline alone tells you the framing: one billionaire's emotional decision-making is the story.

That framing misses the bigger picture.

The real concern isn't whether Son is "starstruck." The concern is whether any of the capital flooding into OpenAI — from SoftBank, Microsoft, or elsewhere — can be justified by actual returns before the IPO window closes.

OpenAI is burning cash at a pace that would be criminal in any other sector of the economy. The Guaranteed Capacity program is savvy, but it also signals that the company needs committed revenue badly enough to offer multi-year discounts to lock it in now.

Altman keeps saying OpenAI expects "hundreds of billions in sales by 2030." The company's current annual revenue is a fraction of that target. These projections need results to back them up.

The Supply Chain Attack Nobody's Talking About

Buried in OpenAI's own news feed: a May 13 post about "Our response to the TanStack npm supply chain attack."

OpenAI was caught up in a software supply chain attack targeting the TanStack npm package — a widely used JavaScript library. This kind of vulnerability can expose enterprise customers who build on OpenAI's infrastructure.

Mainstream tech coverage largely skipped this. As OpenAI pushes deeper into enterprise environments — especially with the Dell on-premises deal — security vulnerabilities deserve scrutiny. CISOs deciding whether to trust OpenAI with sensitive workloads need visibility into these incidents.

The Bottom Line

OpenAI is executing aggressively on multiple fronts simultaneously: locking in enterprise compute revenue, expanding into on-premises deployments, and shipping new products at breakneck pace.

The business logic is sound. The financial math is still unproven.

Regular people and businesses betting on OpenAI's ecosystem — signing multi-year Guaranteed Capacity contracts, building products on its APIs, integrating Codex into their workflows — are making a calculated bet that the company's revenue will eventually match its ambition.

If it does, these are smart moves made early. If it doesn't, the fallout from an $850 billion valuation collapse won't stay on Wall Street. It'll land on every developer, enterprise customer, and startup that built their future on OpenAI's promises.