NextFin News - Jensen Huang of Nvidia and Dario Amodei of Anthropic will sit down for lunch on Tuesday with President Donald Trump and House Speaker Mike Johnson to discuss artificial-intelligence risks, setting a rare face-to-face test of whether the executives now calling for brakes on AI development can persuade a president who has publicly called those same fears a "hoax."
The guest list, confirmed by people familiar with the matter, also includes Palantir Technologies CEO Alex Karp and OpenAI President Greg Brockman. The meeting comes two weeks after Amodei published a 3,800-word essay urging the industry to "slow the pace at which we improve the capabilities of AI models," and two weeks after Trump wrote on social media that "AI taking over the World, destroying Humanity, and all other things bad, is a HOAX."
The Lunch and the Contradiction
The core tension is simple: the two most prominent voices for slowing AI are dining with the most prominent skeptic of slowing AI.
House Speaker Mike Johnson framed the session on Monday as a search for balance rather than a prelude to new rules.
"We do not need a moratorium. We do not need to jump in and hyper-regulate this, because we'll lose the race to China. The innovation is important to continue, but we have to strike the right balance. That's what the conversation is going to be about."
That framing leaves little room for what Amodei and a growing group of lab leaders are asking for. Anthropic and OpenAI are both reportedly preparing for initial public offerings that could set records, which gives the safety push an obvious timing problem: the executives who stand to gain the most from the current boom are the same ones asking to slow it down.
Trump has been explicit about where he stands. On September 14, he wrote that "there is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China," and added that "the only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT." He has also compared recent calls for the industry to slow development to politically driven "hoaxes."
Yet the president agreed to the lunch anyway. He also had dinner with Amodei on Sunday, according to reporting on the run-up to the meeting. The sequence suggests the administration wants to hear the safety case without committing to the policy changes that would come with it.
That is the pattern of a regime that governs by access rather than by rule. A meeting costs the president nothing, but it gives the companies a reason to believe they were heard — and gives the White House a record of having listened if the politics turn.
What the CEOs Are Actually Asking For
Amodei's proposal is not the blanket six-month freeze that the Future of Life Institute circulated in 2023. That letter asked for a pause with no enforcement mechanism, and no major lab stopped training. Amodei's framework is more specific and more operational: embedded independent evaluators inside frontier labs, cross-company safety checkpoints, and narrow international coordination on the highest-risk capabilities. He calls it "pacing the frontier" — slowing the rate of capability jumps rather than halting work.
"Progress will still seem fast, and we must make wise use of the time we gain."
The shift from protest to policy draft is why the request is being treated more seriously in Washington than the 2023 pause letters ever were. It also creates a political opening for a Congress that has so far passed no bipartisan AI safety legislation.
House Democratic Leader Hakeem Jeffries said industry leaders were pleading for government action.
"So, when it comes to AI safety, it certainly seems to me that we need to actually lean in boldly and responsibly — now."
But the ask carries a credibility problem that no amount of policy drafting solves. David Sacks, who served as White House AI and crypto czar from January 2025 through March 2026, told the frontier labs they could slow down if they genuinely believed their unreleased models were dangerous — but warned that demanding regulation in exchange could amount to "blackmail of the public and the political system." His point was that the two companies asking for brakes also happen to hold a duopoly on frontier intelligence, and that slowing development locks in their lead.
"Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you'll buy goodwill for the next conversation. If you don't, we'll know this was just another bid for regulatory capture — or an election-season psyop."
Michael Burry, the investor known for betting against the housing bubble, put the financial version of the same argument more bluntly: the slowdown calls are "cover for real uncontrollable slowing growth" as AI labs struggle to maintain the stratospheric expansion of the boom's early phase, and a cooling period would help "as IPOs look to be pushed out."
The Regulatory Baseline: Two Years of Deregulation
Any discussion of new AI rules has to start from where the Trump administration actually stands, and the record is unambiguous. On January 23, 2025, the White House signed Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence." Section 3 revoked President Biden's Executive Order 14110 — the October 2023 order that had used the Defense Production Act to require developers of the largest dual-use foundation models, defined as systems trained with more than 10^26 floating-point operations, to report training activities, model-weight security, and safety red-team results to the federal government. That reporting requirement ended with the revocation, and nothing replaced it.
A second order, Executive Order 14365, signed December 11, 2025, is aimed not at companies at all but at states. It directs the Attorney General to create an AI Litigation Task Force to challenge state AI laws on federal-preemption grounds, and conditions certain federal funding — including Broadband Equity, Access, and Deployment program money — on states dropping what the order calls "onerous" AI laws.
The policy blueprint behind both orders is "Winning the AI Race: America's AI Action Plan," released July 23, 2025, which set out more than 90 federal actions built around deregulation, data-center infrastructure, workforce development, and U.S. dominance rather than mandatory safety standards.
In other words, the CEOs are asking for a new federal safety architecture on terrain the administration has spent 20 months dismantling. The ask is not a tweak to an existing regime; it is a request to reverse a regime.
That asymmetry explains why the lunch matters even if it produces nothing. In the Biden era, AI governance was channeled through agencies, notice-and-comment rulemaking, and statutory reporting requirements — slow, predictable, and legible to markets. Under the current framework, the operative guardrail is personal: the president's own judgment, exercised through executive orders and, most recently, a pledge to form an "AI Force" with a hinted law-enforcement role.
Personal discretion is faster than legislation, but it is also less predictable. A policy that can be announced in a social-media post can be reversed in one. For Nvidia and the labs, the risk is not a specific rule; it is that the rules become a function of access, timing, and presidential mood rather than published criteria.
Nvidia's Asymmetric Exposure
Nvidia sits in an awkward position at this lunch. The company is not a frontier lab; it sells the compute that frontier labs buy. Its exposure to AI policy is indirect but enormous, and it runs in both directions at once.
Nvidia reported revenue of $96.2 billion for the fiscal second quarter ended July 26, up 106% from a year earlier, with adjusted earnings of $2.22 a share. The company estimates it holds roughly 80% of the AI compute market. Its shares closed at $228.86 on September 28, down 0.39% on the day, after touching a 52-week high of $236.54 in mid-May. The company's market value stands around $5 trillion.
Every dollar of that valuation depends on frontier labs continuing to spend on training runs. If the industry genuinely slows the pace of capability improvement, demand for the most advanced chips could cool. If the industry does not slow and a catastrophic failure occurs, the political backlash could produce rules far harsher than anything Amodei is proposing. Nvidia's optimal outcome is the one no one at the table will say out loud: continued breakneck spending, with no catastrophe and no regulation.
The company's board appears to understand the political risk. On Monday, it approved a $150 billion increase to its share-repurchase program, bringing total authorization to $235 billion — the largest buyback in U.S. corporate history. Cash is cheap insurance when your biggest customers may soon be asking regulators to tell them to build fewer of your products. The buyback also signals that management sees limited upside in further growth at current valuations, or at least wants to return capital before the policy picture clarifies.
The second-order transmission chain is where the lunch reaches investors who never touch an AI stock. Policy uncertainty raises the discount rate the market applies to long-duration growth assets. A higher discount rate compresses the present value of earnings expected years out — exactly the earnings profile of a company whose valuation rests on AI infrastructure spending through the end of the decade. So a lunch about AI safety becomes, through the discount-rate channel, a story about the entire technology complex. That is why the market watches these meetings even when the agenda is vague.
The Counter-Thesis: This Lunch Changes Nothing
The strongest case against reading too much into Tuesday is that the administration's position is already locked in. Sacks, the former AI czar, told the frontier labs they could slow down on their own but should not demand regulation in exchange. Vice President JD Vance made the same point more bluntly.
"I feel a little bit weird about the fact that you have so many frontier AI tech companies kind of coming to the government and begging the government to regulate them."
On this reading, the lunch is optics. The executive orders stand. No bipartisan bill is moving. The midterm message is about winning the race against China, not about safety guardrails. The CEOs' ask is public-relations positioning ahead of IPOs, and the administration is happy to take the meeting and give nothing back.
That case is strong on the record. But it misses the mechanism that matters. Personal discretion cuts both ways: a president who dismissed AI fears as a hoax in September can still sign an order in October if a high-profile failure forces his hand. The absence of a statutory framework does not eliminate regulation; it concentrates it in the executive, where the response time is shorter and the political incentive to be seen acting is stronger.
The 2023 pause letters failed because they asked for a moratorium with no mechanism and no constituency. This round is different because the ask is narrower, the companies are closer to public markets, and the administration has already shown it will act by decree when it wants to. The question is not whether the CEOs get what they want. It is whether the administration decides it wants something they can give.
The Cyclical Read and the Structural Shift
The lunch itself is cyclical: a sentiment event, a photo opportunity, a data point in the election-year messaging cycle. Sentiment reverts. If nothing concrete follows, the market will forget it within a quarter, and the policy baseline will look exactly as it does today. That is the mean-reverting leg, and it argues for not over-reading Tuesday.
But the structural leg sits underneath it, and it does not revert on its own. The United States has replaced a rule-based AI governance regime — notice-and-comment rulemaking, statutory reporting thresholds, agency processes with published timelines — with a personal one, in which the operative constraint is a president's judgment and the operative instrument is an executive order or a social-media post. A structural shift is defined by a change in the rules of the game that does not self-correct, and this is one: the guardrail moved from the statute book to a person. That divergence from the European Union's rule-based AI Act is not a cycle; it is a regime difference, and it will price into cross-border capital allocation for years.
The investment implication of getting this call wrong is asymmetric. If you treat a structural shift as cyclical, you buy the dip every time a post rattles the market, and you keep buying until the regime change reprices the entire complex. If you treat a cyclical event as structural, you sit out a rally that resumes on schedule. The evidence here — revoked reporting requirements, a litigation task force aimed at state laws, funding conditioned on deregulation, and a president who calls safety fears a hoax — points to structure. The lunch is noise on top of a signal.
Scenarios: What Priced In Looks Like
The base case, and the one the market has effectively priced, is theater: the lunch happens, the readouts emphasize balance and the China race, no framework lands within 30 days, and Nvidia trades on earnings and capex rather than policy. The trigger that would break this base case is a named AI Force leader with a law-enforcement mandate announced before the midterms.
The upside case for the deregulation trade is that the administration converts the lunch into a legislative vehicle that codifies the current hands-off posture before November, locking in the policy baseline beyond the next administration. That would remove the discount-rate overhang on long-duration AI assets and re-rate the entire infrastructure complex.
The downside case is that a high-profile AI failure — a breach, an autonomous-agent incident, a fatality tied to a model — forces the executive to act by decree, producing the very rules the CEOs asked for but in a harsher, faster, less predictable form than the negotiated version. That is the path where the lunch looks, in hindsight, like the last polite conversation before the regime changed its mind.
What to Watch
Three signals will tell whether Tuesday produces substance or theater. First, whether the White House announces a concrete framework for frontier-model evaluation within 30 days — an executive order, a legislative proposal, or a formal Justice Department inquiry. Second, whether the "AI Force" pledge takes institutional form with a named leader and a defined mandate. Third, whether Anthropic and OpenAI actually file for their IPOs in the coming months; a filing soon after the safety push would feed the regulatory-capture narrative, while a delay would suggest the slowdown ask is more than positioning.
Short-term, expect volatility around any hint of enforcement without clarity — markets dislike uncertainty more than they dislike rules. Medium-term, the question is whether the administration converts the lunch into a legislative vehicle before the November midterms. Long-term, the structural shift is the divergence between the U.S. approach — personal, discretionary, dominance-first — and the European Union's rule-based AI Act. That divergence itself becomes a cost of doing business for any company operating on both sides of the Atlantic, and a source of arbitrage for any company that can choose its jurisdiction.
The falsifying signal for the view that this matters: if 60 days pass with no new framework, no AI Force appointment, and no change in the enforcement posture, then the lunch was theater, and the deregulatory baseline holds.
The deeper story is not whether AI gets regulated. It is who decides, and by what process. A guardrail that lives in one president's judgment is not a guardrail — it is a mood, and moods change faster than chips depreciate.
Explore more exclusive insights at nextfin.ai.
