Jul 9, 2026, 10:17 AM ETOther

Enterprises Fear Frontier Models | Sam Altman Offers Trump 5% of OpenAI | DeepSeek Builds Own Chips

The U.S. government has lifted a 19-day ban on running Fable models, though this has not eliminated the need for a structured pre-approval process for future deployments.
Sam Altman (OpenAI) floated a proposal for the U.S. government to take a 5% equity stake in OpenAI to align interests and mitigate regulatory friction.
OpenAI previously released a nine-point plan suggesting a restructuring of the U.S. taxation system to tax capital gains more and income less due to AI-induced job displacement fears.
Speakers characterize the 5% stake proposal as a strategic "anchor" to prevent the government from demanding a larger (e.g., 50%) ownership share later.
One speaker argues that government ownership does not guarantee alignment, citing the TARP bailout where government stakes did not prevent political interference.
Another perspective suggests that giving a small stake creates an "unexpectedly large" amount of alignment by bringing the government into the boardroom dynamic.
The discussion highlights a paradigm shift from the 1990s deregulation era (Telecom Act, Section 230, no sales tax) to a current era where tech companies are voluntarily seeking government regulation and ownership.
The "pre-approval" process for AI shipping is viewed as a major structural change, potentially making the U.S. economy less dynamic compared to the lack of pre-approval in the past.

Meta launched a cloud computing business to sell excess AI infrastructure, causing its stock to jump 10% (its biggest single-day gain in five months).
SpaceX and Meta have both pivoted from building proprietary AI models to selling excess compute, a move the market views positively.
The entry of Meta and SpaceX into the neo-cloud market caused stock prices of pure-play competitors (CoreWeave, Nebius) to drop by approximately 10%.
NVIDIA is introducing "Compute Now, Pay Later" financing, offering backstop rights and revenue-sharing credit to neo-cloud providers to absorb hardware risk.
DeepSeek has announced plans to develop its own AI chips, signaling a trend of major AI players moving toward vertical integration.
Anthropic has opened talks with Samsung to build proprietary AI chips, following the same vertical integration logic.
Cling (a Chinese AI video company) raised $2.8 billion at an $18 billion valuation with $500 million ARR, suggesting a potential AI valuation bubble in China or higher risk appetite.
OpenAI shut down its consumer Sora product due to cost inefficiencies, whereas Cling successfully monetizes at a scale that generates profit.
The "app layer" is facing pressure to contain costs per resolution, with the CX industry standardizing around $0.50 per resolution, forcing a push toward more efficient models.

Microsoft and Amazon are deploying thousands of engineers (FDEs) inside enterprise clients to drive AI adoption, aiming to replicate the "services" model of the IBM era.
Jason argues this enterprise services strategy will largely fail due to a critical shortage of talent capable of handling complex AI deployments and change management.
A specific case study cited involved a client waiting three months to fix a bug while an FDE was on paternity leave, illustrating the lack of depth in current technical teams.
The rate of AI technology diffusion is expected to slow down in the enterprise sector due to these adoption bottlenecks, potentially taking 3x longer to achieve the next 10x growth.
Successful enterprise deployments increasingly require a "three-way team" comprising a technical expert, a domain expert (e.g., lawyer, industry specialist), and the vendor.

The Venture Capital industry is experiencing a normalization of founder insensitivity to dilution, with many founders accepting multiple high-dilution rounds (e.g., 20+ rounds) to avoid valuation traps.
Late-stage investors in mature companies are no longer blocking exits over 1x returns, reducing friction for founders seeking to sell.
The "high-priced round" fear has diminished; founders now accept that exiting with a partial return is acceptable if it allows them to stay in the game for a potential massive outcome.
Investor Ashton Kutcher announced his departure from Sound Ventures to launch a new firm with Morgan Bella, moving from late-stage investments to deep tech seed.
11 Labs is trading at a secondary market valuation of $22 billion, highlighting a market for secondary liquidity in private companies.
A key debate emerged regarding employees joining startups: the consensus is that talent should join companies with a high probability of launching tender offers (liquidity events) within 1–3 years to mitigate the risk of illiquid equity.
Some top-tier startups are removing vesting cliffs entirely for senior employees to simplify liquidity and retention.

China is rapidly advancing its own frontier AI models (e.g., Kling, open-source alternatives) in response to U.S. export controls that block access to tools like Claude and GPT.
Chinese open-source models now dominate the top six rankings on OpenRouter, while U.S. models lead in coding and closed-source enterprise applications.
The U.S. decision to restrict AI chip and model access to China is expected to force China to build competitive domestic alternatives, accelerating the global split in AI capabilities.
There is a growing trend of companies moving toward open-source models to manage costs and avoid vendor lock-in, driven by the diminishing marginal returns of cheaper models for complex tasks.
For highly complex problems, enterprises are still forced to use frontier models (e.g., Fable, Opus) despite higher costs, as open-source models lack the necessary resolution for "unknown unknowns."