TL;DR
- Manufacturing push: OpenAI launched a new Request for Proposals (RFP) for US manufacturers to supply components for data centers, robotics, and consumer devices by June 2026.
- Compute diversification: The company will integrate 750 megawatts of Cerebras compute in a $10 billion deal, reducing Nvidia dependence.
- Consumer hardware: OpenAI is developing ‘Sweetpea’, a Jony Ive-designed audio wearable targeting 40-50 million units, launching September 2026.
- Robotics return: The company is re-entering robotics after abandoning it in 2021, hiring engineers and exploring humanoid form factors.
- Infrastructure scale: Total infrastructure investment exceeds $1.4 trillion for 30 gigawatts of compute capacity through 2028.
OpenAI on Wednesday announced a request for proposals from US-based manufacturers for components related to data centers, robotics, and consumer devices, with submissions due June 2026.
Beyond data centers, the move signals a hardware transformation encompassing robotics and consumer electronics, including an audio wearable designed by Jony Ive scheduled for September 2026 with an aggressive 40 to 50 million unit target.
“AI is a catalyst for the reindustrializing of the country,” declared Chris Lehane, OpenAI’s Chief Global Affairs Officer. “We do have to bring back the supply chain here.” OpenAI positions the RFP as a “historic opportunity to strengthen domestic supply chains and reindustrialize the country,” aligning with Trump administration manufacturing priorities while addressing fundamental constraints in AI infrastructure.
Promo
Compute Infrastructure at Trillion-Dollar Scale
Underpinning this manufacturing transformation is an infrastructure investment exceeding a trillion dollars in total cost. OpenAI will integrate 750 megawatts of ultra-low latency compute from chipmaker Cerebras in a deal worth over $10 billion, running from 2026 through 2028.
Overall, the company committed to about 30 gigawatts of compute with a total cost of ownership of about $1.4 trillion.
“OpenAI’s compute strategy is to build a resilient portfolio that matches the right systems to the right workloads,” explained Sachin Katti of OpenAI in a recent announcement of a new partnership with infrastructure provider Cerebras.
What drives this diversification strategy? When all AI companies depend on the same supplier for components, hardware access becomes the bottleneck determining who can scale. OpenAI views diversification as existential because supplier dependence eliminates competitive differentiation.
By securing alternative compute sources like Cerebras, OpenAI aims to control enough infrastructure to negotiate from strength. In October 2025, Sam Altman noted the company would like to build an AI factory to make 1 gigawatt of compute per week, ambitions that dwarf current industry production rates.
Direct Challenge to Apple’s Hardware Dominance
Beyond data center infrastructure, OpenAI’s hardware push extends to consumer devices that directly challenge established players. OpenAI is improving its audio AI models and expects users of future personal AI devices to communicate through voice commands rather than via a screen.
A new audio model is scheduled for release during the first quarter of 2026, followed by its first personal AI device in early 2027.
OpenAI is developing ‘Sweetpea’, a premium audio wearable scheduled for September 2026 launch that would compete with Apple’s AirPods. Features include a custom processor and advanced voice AI capabilities enabling users to control phones and execute complex tasks through natural conversation.
According to reports, OpenAI is targeting an aggressive 40 to 50 million units of Sweetpea in the first year.
Such production targets reveal a fundamental strategic shift. Targeting tens of millions of units in year one means OpenAI must secure manufacturing capacity, component suppliers, and quality control systems comparable to established consumer electronics giants.
OpenAI believes its AI advantage can overcome Apple’s decade-long head start in audio wearables, betting that superior voice capabilities will matter more than hardware refinement or ecosystem integration.
Personnel behind Sweetpea amplifies the competitive threat. Jony Ive led the design team for the iMac, iPhone, and iPad and is now working on a device with OpenAI.
OpenAI last year merged with io, an AI device startup co-founded by Ive, representing a direct challenge to Apple in a market it has dominated since 2016. Up to five devices are planned by late 2028.
Robotics Re-entry After 2021 Abandonment
While OpenAI advances on consumer electronics, it is simultaneously reviving robotics after abandoning the field in 2021. OpenAI has brought on Meta’s AR leader Caitlin Kalinowski and is also exploring humanoid robots.
What changed to enable this strategic reversal? Limitations in training simulated agents for real-world tasks drove the 2021 exit. OpenAI’s current foundation models now demonstrate multimodal capabilities that can process vision, language, and physical control simultaneously, indicating the technological gap that forced the earlier retreat has closed.
OpenAI has also driven investments in robotics startups like 1X Technologies and Figure building relationships in a sector that raised $3.2bn globally in 2025, exceeding the previous six years combined.
Implications of this dual approach point to strategic hedging. By investing in external robotics companies while developing internal capabilities, OpenAI gains access to humanoid robots whether through portfolio companies or proprietary development.
Mirroring the Cerebras compute strategy, OpenAI is securing optionality across the value chain. A consistent philosophy emerges: when facing technological uncertainty, control multiple pathways rather than committing to a single bet.
Supply Chain Strategy: Nvidia Dependence and Bottlenecks
Across all these hardware initiatives – data centers, consumer devices, robotics – a common challenge emerges: securing reliable supply chains.
While the Request for Proposals requests don’t indicate how much OpenAI intends to spend or over what time frame, the scope signals a strategic shift toward vertical integration.
“When people talk about AI infrastructure, the conversation often stops at chips and data centers,” says OpenAI. “But advanced AI depends on a much broader ecosystem of physical components.”
OpenAI detailed the RFP in correspondence with the Office of Science and Technology Policy (OSTP), describing it as a “once-in-a-century opportunity to strengthen our economy and modernize America’s industrial base.”
Despite ambitious rhetoric, concrete constraints loom. OpenAI estimates their five-year buildout will strain existing skilled trade workforces by 20%. National targets now include building 100 GW of new energy capacity annually.
Lead times for specialized equipment are stretching past two years. OpenAI and competitors are almost exclusively dependent on Nvidia’s supply chain. High Bandwidth Memory (HBM) supply is fully committed for 2026, with tight conditions expected beyond that year.
Supply chain dependence creates a strategic paradox for AI companies. When every competitor sources from identical suppliers, technical superiority in AI models matters less if you cannot secure enough hardware to deploy them at scale.
Companies with superior supply chain control can outcompete rivals not through better algorithms but through sheer deployment capacity. OpenAI views vertical integration as existential rather than optional for this reason.
Google leverages its own specialized TPUs giving them substantial hardware autonomy, while OpenAI remains dependent on external suppliers.
