Sam Altman once floated a clean break. Late last year the OpenAI chief executive sketched out a plan to spin off the company’s robotics and consumer-hardware units. The goal looked straightforward on paper. Let those capital-hungry teams raise money on their own. Shield the core artificial-intelligence business from losses that could spook public-market investors. Give the divisions room to operate like internal startups.
But accountants delivered the verdict. Even as separate companies the units would likely stay consolidated on OpenAI’s balance sheet. The proposal died. No active talks continue today. Yet the episode reveals how physical-world ambitions now test the economics that once made OpenAI the clearest bet in technology.
The robotics and hardware teams already function at arm’s length. They report directly to Altman. Employees sometimes describe them as separate companies housed inside the larger organization. That separation once fueled optimism. It also highlighted the tension. Software scales with remarkable margins. Hardware demands years of spending before revenue appears.
OpenAI paid $6.5 billion in stock last May to buy io, the AI hardware outfit led by former Apple designer Jony Ive. The deal brought roughly 55 employees and a vision for a new class of device. Altman described it as something fully aware of its user’s surroundings, small enough for a pocket, the third object a person would set on a desk after a MacBook Pro and an iPhone. Legal filings show customers won’t see it before the end of February 2027. That timeline stretches beyond any near-term IPO window.
The Accounting Wall That Stopped the Spin-Off
Altman’s idea echoed Google’s 2015 restructuring into Alphabet. Core search would stand apart from bets like self-driving cars and life sciences. OpenAI weighed a similar split for robotics and hardware. The divisions could chase outside capital. They could move faster without dragging down the profit-focused AI engine.
But accounting standards refused to cooperate. The Wall Street Journal first reported the proposal and its rejection, citing people familiar with the matter. Even independent entities would remain tied to OpenAI’s financials if the parent retained control or significant influence. The hoped-for clean sheet never materialized.
This matters now because OpenAI races toward a public debut. The company closed a $122 billion funding round earlier this year, the largest in Silicon Valley history. That round valued it at $852 billion post-money. Investors bet on explosive AI demand. They also bet that OpenAI can translate that demand into sustainable profits before the IPO. Recent internal misses on revenue and user targets have sharpened the focus.
So OpenAI cut side projects. It discontinued the standalone version of its video-generation tool Sora to free computing power for higher-priority work. The company pivots toward a new superapp aimed at coders and enterprise customers after falling behind rival Anthropic in some metrics. CFO Sarah Friar has scrutinized spending with fresh intensity.
And still the hardware bets continue. OpenAI has worked on robotics for years. Early experiments trained a humanlike hand to solve a Rubik’s Cube. Last year the company announced a research partnership with Coco Robotics, a delivery robot service where Altman also invests personally. Altman told the Core Memory podcast last month, “We’re trying to figure out how to be very successful at robotics. If you could pick one thing to make the U.S. competitive at manufacturing and the world of atoms in general, you would say we need a lot of robots that can build a lot, lot more robots.”
That conviction runs deep. But execution carries real costs. The io acquisition alone exceeds annual spending at some hardware rivals. Early-stage hardware businesses often burn cash for years. Tesla’s automotive and energy units once did the same. Public markets forgave those losses because the vision stayed clear and execution eventually improved. OpenAI must now prove the same path exists for AI embodied in metal and silicon.
History inside OpenAI adds caution. The company disbanded its original robotics research team years ago after running short on data to train models on physical tasks. It quietly restarted the effort in 2024, hiring specialists in humanoid control and teleoperation. By early 2026 the unit’s leader, Caitlin Kalinowski, had departed. She raised concerns about a rushed partnership with the U.S. Department of Defense involving surveillance and lethal autonomy questions. Altman later acknowledged the haste, saying the company “shouldn’t have rushed to get this out on Friday.”
These episodes show the complexity. Building machines that perceive, act and learn in unstructured environments requires massive data, compute and capital. OpenAI repurposes technology from Sora to simulate worlds for robot training. It hunts for U.S. suppliers to build secure supply chains. It develops custom sensors. All of it costs money today. Revenue sits years away.
Why Hardware Complicates the Public Market Story
Public investors prize predictability. Software margins often exceed 80 percent once scale arrives. Hardware margins arrive later and usually land lower. The $6.5 billion io deal, the delayed 2027 device launch, the open-ended robotics research, these items create a different risk profile than ChatGPT subscriptions or enterprise API calls.
Analysts will scrutinize the S-1 filing when it arrives. They will look for line items that show how much cash flows into physical bets versus core AI. The scrapped spin-off attempt already signals that Altman once wanted those losses isolated. The fact that accounting rules blocked the move means they will sit inside the main numbers.
Yet OpenAI holds advantages. Its latest funding round brought in Amazon, Nvidia, SoftBank and others. That capital buys time. Enterprise revenue now makes up more than 40 percent of total revenue and heads toward parity with consumer. New models drive record engagement in agentic workflows. The company projects continued growth even after recent target misses.
But the hardware question lingers. Success in physical AI could unlock trillions in economic value through smarter manufacturing, logistics and personal assistance. Failure would weigh on margins for years. Altman has described robotics as essential for national competitiveness in the world of atoms. That belief explains why the divisions survived the recent round of project cuts.
Recent coverage reinforces the tension. Yahoo Finance detailed how the spin-off proposal collapsed over balance-sheet concerns. The Verge noted that OpenAI has cut back on side quests ahead of the IPO but could revive spin-out ideas later. On X, investors and analysts pointed to projected $14 billion losses for 2026 against $13.1 billion in revenue last year. One post observed that pre-IPO leaders rarely try to hide loss-making divisions unless those losses threaten the core narrative.
OpenAI has restructured to give safety and security clearer lines of authority. Chief Research Officer Mark Chen and President Greg Brockman now hold distinct roles. Altman focuses more on capital raising, data centers and supply chains. The product organization reframes its work as AGI deployment across both digital and physical domains.
The original WebProNews analysis captured the stakes well. Hardware ambitions require patience that public markets sometimes withhold. They also demand different operational skills than training ever-larger language models. Supply chains, hardware engineering talent, regulatory questions around autonomous systems. All add layers of execution risk.
So far investors have rewarded the vision. The $852 billion valuation reflects belief that OpenAI sits at the center of the AI boom. But valuations compress when growth slows or costs accelerate. The robotics and hardware units must either deliver measurable progress or risk becoming the question every analyst asks on earnings calls.
Altman once said he was zero percent excited to become a public-company CEO but saw some upside for OpenAI going public. The scrapped spin-off suggests he understands the trade-offs. Clean financials help IPO pricing. Continued heavy investment in physical AI could deliver the next leap in capability. Reconciling both tests the organization’s discipline in the months ahead.
The hardware road remains open. Partnerships with existing robot makers could accelerate deployment without full ownership of manufacturing. Internal research may yield breakthroughs that justify the spend. Yet the accounting reality that killed the spin-off plan also sets the rules for how Wall Street will judge the results. OpenAI no longer operates only in the world of bits. It has stepped into atoms. The balance sheet will show the cost of that step long before the first robot walks out of a factory or the pocket device ships to customers.
