OpenAI floats buy-before-your-try AI availability guarantee

OpenAI is now offering its customers a guarantee that it will actually provide the service it's selling if customers need more AI sauce. All that's required is an annual spending commitment. AI is in short supply as demand - stoked by flat-rate subscriptions - races ahead of datacenter inference capacity. AI workloads require a lot of computing power, especially when they run for hours on end, a common scenario for AI agents. They require more computing power than OpenAI and its cloud partners can provide. Promised datacenter construction is unlikely to address this shortfall any time soon, assuming those projects are actually completed at all. The result has been thinly disguised rationing through usage limits, variable pricing, subscription restrictions, and token consumption arbitrage - swapping out expensive models for cheaper ones. OpenAI's answer is a new offering meant to ensure it can deliver on its AI promises. "Today we announced OpenAI Guaranteed Capacity, a new offering that helps eligible customers plan for reliable access to OpenAI compute across supported cloud providers as they scale critical workflows," said Sachin Katti, who runs compute for OpenAI, in a LinkedIn post. "It gives customers a clearer framework to align forecasted demand, commercial commitments, and guaranteed shared capacity over time." According to Katti, AI adoption has made processor availability the salient constraint on modern software. He argues that companies dependent on AI workflows need to take steps to ensure infrastructure availability. Guaranteed Capacity is intended to ensure that if client demand grows, compute infrastructure will be available to support that growth. OpenAI says customers can make annual spending commitments ranging from one to three years, with discounts that scale according to duration. "Guaranteed Capacity includes certainty of access to compute based on spend levels, and customers can draw down from this commitment across the portfolio of OpenAI products," the company says. Not everyone is buying OpenAI's pitch. Santosh Ahuja, founder and CEO of enterprise infrastructure biz Pervasiviti and an experienced CTO and researcher, challenged Katti's announcement as a repackaging of basic cloud service obligations in a discussion thread reply. "Every hyperscaler solved 'reserve now, scale later' a decade ago," he wrote. "So the actual announcement is: OpenAI now offers what every cloud provider has treated as a basic procurement primitive but positions it as a strategic differentiator." Ahuja argues that OpenAI's AI evangelism can't keep up with its execution. "Enterprises don't need a 'clearer framework to align forecasted demand,'" he wrote. "They need deterministic SLAs with penalty clauses." In a message to The Register, Ahuja explained that the problem is not guaranteed capacity, it's the way OpenAI has framed cloud business assumptions as innovation. "Reserved instance models have existed at every hyperscaler for over a decade - AWS, GCP, Azure," he explained. "When you strip the marketing, OpenAI is acknowledging that demand has outrun their infrastructure planning and asking customers to pay for predictability that mature platforms deliver by default. "The real question enterprises should ask is whether these guarantees come with enforceable SLAs and financial penalties for breach, because without those, 'guaranteed' is just a word in a press release." Ahuja pointed to a response to Katti's post from Keith Strier, SVP of global AI markets at AMD ("Brilliant. We accept the challenge.") which he takes to mean, "We will try our best." "That's your supply chain partner speaking - and 'we will try our best' is not the language of guaranteed anything," Ahuja said. "It sounds like OpenAI is promising capacity they don't yet have firm commitments on from their own silicon suppliers. That's not a product launch. That's wishful thinking with a price tag." So too is OpenAI's preparation to go public, which reportedly could result in a corporate filing as soon as this week. (R)