OpenAI PreStocks (OPENAI) on Solana

Sam Altman has invested hundreds of millions of dollars in Oklo Inc. (NYSE: OKLO), a nuclear startup building small modular reactors (SMRs) designed to power AI data centers directly without routing through the traditional electric grid. The strategic logic is straightforward: training and running large language models demands enormous amounts of electricity, and nuclear generation can supply that power without increasing residential bills or competing with public utility demand. Oklo is already in commercial discussions with customers including Meta Platforms, positioning itself as a dedicated energy provider for the AI industry.

Oklo remains pre-revenue with negative free cash flow of $154 million over the past twelve months, and its reactor design has not yet received full Nuclear Regulatory Commission approval. The company is working with the Department of Energy on a pilot reactor in Idaho, but analysts estimate commercial operation is potentially a decade away. Altman's bet reflects a broader thesis that OpenAI and peer AI labs will need to secure dedicated, long-term energy supply chains independent of the public grid as compute demands continue to scale — even if the path to operational nuclear capacity is measured in years rather than quarters.

Elon Musk said in an interview with The Economist that OpenAI has become "the exact opposite" of what he envisioned when he co-founded it in 2015. Musk, who donated less than $45 million before departing in 2018, argued that the company's original mission — to serve as an open-source counterweight to Google — has been abandoned in favor of a closed-source, for-profit structure. OpenAI completed its restructuring into a public benefit corporation in October 2025, with Microsoft holding a 27% stake and the company now valued at over $850 billion ahead of a potential IPO.

Musk framed his criticism as concern over mission drift rather than personal animosity, though he is also pursuing a separate lawsuit against the company. He pointed to Anthropic — founded by former OpenAI researchers — as evidence of broader trust issues inside the lab, praising Anthropic CEO Dario Amodei as principled. Despite his grievances, Musk said he remains open to dialogue with OpenAI leadership on AI safety. The remarks reinforce a continuing narrative around OpenAI's commercialization trajectory and the governance tensions that have followed its rapid valuation growth.

OpenAI CFO Sarah Friar has outlined a new framework for evaluating AI investments centered on a "Useful Intelligence per Dollar" metric, shifting from traditional adoption measures like licenses and active users to outcome-based performance. The framework asks four core questions: whether AI completes meaningful work, the cost per successful task, result reliability, and how value scales with usage.

To calculate cost per successful task, Friar recommends dividing total completion costs—including employee time, human review, retries, and rework—by the number of tasks meeting quality standards. She noted that "the lowest price per token does not always produce the lowest cost per outcome," signaling OpenAI's intent to help enterprises quantify real business value as AI moves from experimentation into production workflows.

The Trump administration has moved to assert direct control over which companies and entities can access frontier AI models, according to CNBC reporting based on sources familiar with the matter. Executive Order 14409, signed June 2, 2026, requires frontier AI developers including OpenAI to give the federal government up to 30 days of early access before models are released to trusted partners. While the White House frames the arrangement as voluntary and not a preclearance requirement, OpenAI's experience with its GPT-5.6 release illustrates the practical impact: the company staggered the rollout following a federal security review request, announcing only a limited preview for trusted partners on June 26 while coordinating with the administration before a broader release.

The policy represents a structural shift in who controls distribution of the most capable AI systems. Historically OpenAI decided which enterprises and agencies received early or priority access to its latest models; under the new framework, that gatekeeping function has been partially transferred to the White House. Frontier model teams at OpenAI now face expanded pre-release requirements — including cyber evaluations, partner-access criteria, government documentation, and incident escalation protocols — adding compliance overhead to major launches. Critics note the voluntary framing conflicts with the evident operational reality, as OpenAI's own release cadence has already been shaped by federal coordination.

Mira Murati, who served as OpenAI's Chief Technology Officer before departing in late 2024, this week released the first model from her new startup Thinking Machines Lab. The model, called Inkling, is a 975-billion-parameter mixture-of-experts transformer trained on 45 trillion tokens — but Thinking Machines acknowledges it does not outperform the best available models, open or closed. Notably, Inkling's architecture mirrors that of DeepSeek-V3, and its post-training relied on synthetic data generated by Kimi K2.5, a Chinese open-weight model from Moonshot AI. On key benchmarks including SWE-Bench Verified, Inkling trails Zhipu's GLM 5.2 and Moonshot's Kimi K2.6.

The strategic bet Thinking Machines is making — open-weight licensing under Apache 2.0 paired with enterprise fine-tuning tooling — is less a technical challenge to OpenAI than a structural one. With Chinese models now handling roughly 45 percent of enterprise tokens routed through OpenRouter, and regulatory scrutiny making those models difficult for U.S. government contractors to adopt, domestically-sourced alternatives gain market access regardless of benchmark performance. That Murati, who had direct knowledge of OpenAI's internal capabilities, chose to build on Chinese research architectures is being read as a signal about where open-model technical leadership currently sits.

A Bloomberg report reveals new details about OpenAI's first consumer hardware device: an AI-powered smart speaker equipped with a camera and sensors that the company positions as meaningfully different from existing products like Apple's HomePod. According to the report, the device is designed to become "increasingly personalized and proactive" over time as it builds an understanding of its owner, with capabilities spanning smart home control, media playback, question answering, and full access to ChatGPT. OpenAI's stated goal is to explore "features and form factors that would make the devices unlike anything currently on the market."

The product is a direct output of OpenAI's $6.5 billion acquisition of io Products — a hardware startup co-founded by former Apple design chief Jony Ive — and is seen internally as strategically important ahead of the company's planned IPO. The hardware push has also drawn legal scrutiny: Apple filed suit alleging OpenAI used trade secrets gained through the hiring of Tang Tan, a 25-year Apple veteran now serving as OpenAI's chief hardware officer, along with more than 400 other former Apple employees. OpenAI denied the allegations. No pricing or launch timeline has been disclosed.

OpenAI CEO Sam Altman publicly pushed back on Apple's trade secret lawsuit, posting on X that he is "not afraid of apple, but i have tremendous respect for them" and that OpenAI has "no interest in other companies' trade secrets." Apple filed suit on July 12, 2026 alleging that OpenAI encouraged its employees to share confidential information — including components, drawings, and materials related to unreleased products — as part of OpenAI's own hardware development push. The lawsuit names Tang Tan, OpenAI's chief hardware officer, a 25-year Apple veteran who left in 2024 to join designer Jony Ive's startup LoveFrom before both joined OpenAI following the company's acquisition of that firm.

The dispute marks a sharp deterioration in what was once a close commercial relationship: OpenAI previously powered Apple Intelligence and Siri integrations. The legal fight adds to uncertainty around OpenAI's IPO timeline, which reports now indicate is unlikely before 2027, and comes as Anthropic has edged ahead of OpenAI in valuation at $1.08 trillion versus $868.3 billion.

Fidji Simo, OpenAI's Chief Operating Officer and the company's second-most senior executive, is stepping down after announcing a medical leave in April 2026. Simo — who served as president and right hand to CEO Sam Altman, overseeing the CFO, Chief Revenue Officer, and leading product and business divisions — cited Postural Orthostatic Tachycardia Syndrome (POTS), a chronic neuroimmune condition she has managed since 2019, as the reason for her departure. She will transition to a part-time adviser role.

The exit adds to a run of C-suite turnover at OpenAI: in April 2026, three executives left simultaneously — chief product officer Kevin Weil, Sora head Bill Peebles, and B2B CTO Srinivas Narayanan. The departures come as the company navigates a more difficult IPO path, having pushed its public offering to 2027 amid declining market enthusiasm and intensifying competition from Anthropic, which recently surpassed OpenAI's valuation for the first time. OpenAI has secured a $520 million credit line from Bank of America as it works toward a target valuation above $1 trillion.

Apple filed a trade secret lawsuit against OpenAI on July 10, 2026, in the U.S. District Court for the Northern District of California, alleging a coordinated scheme to steal hardware-related intellectual property that reached "every level" of the company. The complaint centers on Tang Tan, OpenAI's Chief Hardware Officer and a former Apple VP who spent 24 years designing iPhone and Apple Watch hardware, who is accused of soliciting job candidates to bring Apple components to interviews, sharing confidential project code names during recruiting, coaching departing Apple employees on how to evade security procedures, and requesting details about unannounced products. A second named defendant, Chang Liu, allegedly retained a work-issued laptop after leaving Apple for OpenAI in January 2026, accessed a former colleague's Apple computer, and downloaded dozens of confidential hardware files including engineering presentations, technical specifications, and proprietary project data. Apple had sent OpenAI a warning letter in February 2026 that went unanswered before filing suit.

The lawsuit arrives as OpenAI has been building out an ambitious hardware roadmap: the company acquired Jony Ive's design firm io for $6.5 billion in May 2025, and analysts have suggested OpenAI is developing an AI-native smartphone. Apple is asking the court to bar OpenAI from using or disclosing the alleged secrets and to require the return of all confidential materials. OpenAI denied the allegations, stating it has "no interest in other companies' trade secrets."

OpenAI and Cerebras Systems have announced a deal valued at over $20 billion spanning several years, described by Cerebras CEO Andrew Feldman as "one of the largest deals in Silicon Valley history." The agreement focuses on fast AI inference for enterprise applications, with OpenAI's Phi-6 model set to run on Cerebras infrastructure at 750 tokens per second — roughly an order of magnitude faster than comparable alternatives. Sam Altman reportedly initiated the partnership in summer 2025 after identifying inference speed as a critical product constraint.

To support the OpenAI workload and additional enterprise customers, Cerebras is deploying billions of dollars in new European data center capacity, targeting 200 megawatts across sites in Lyon (France), Norway, and Finland by end of 2026. The expansion underscores OpenAI's growing focus on latency-sensitive AI agent applications and its need for diversified, high-throughput compute infrastructure outside of its existing hyperscaler relationships.