• President Trump and House Speaker Mike Johnson meet top AI executives Tuesday to discuss balancing innovation and safety.
  • Attendees include Elon Musk, Jeff Bezos, Satya Nadella, Jensen Huang, Mark Zuckerberg, Lisa Su, Hock Tan, and Dario Amodei.
  • The administration favors industry-led safeguards and existing agency oversight over a new federal AI regulator.

Trump Hosts AI Power Lunch with Tech Titans as Washington Weighs Light-Touch Rules

President Trump and House Speaker Mike Johnson are hosting a high-stakes lunch at the White House on Tuesday with the leaders of America's most powerful technology companies, as Washington grapples with how to oversee rapidly advancing artificial intelligence without stifling innovation.

The guest list, first reported by Axios and confirmed by people familiar with the matter, reads like a who's who of the AI boom: Elon Musk, Jeff Bezos, Satya Nadella, Jensen Huang, Mark Zuckerberg, Lisa Su, Hock Tan, and Dario Amodei. The gathering underscores the administration's preference for an industry-led approach to AI safeguards, even as pressure mounts for government oversight.

Speaker Johnson said the meeting is intended to find a balance between innovation, transparency, oversight, and safety—not to halt AI development. He suggested using existing enforcement tools, including potential Justice Department involvement if harms occur, rather than creating a wholly new AI regulator. "We want to make sure that we're not overregulating, but we also want to make sure that we're doing it right," Johnson told reporters ahead of the lunch.

The meeting follows a private dinner between Trump and Anthropic CEO Dario Amodei on September 27, according to two people familiar with the matter. Amodei, along with Bill Gates and several other frontier-AI leaders, has recently pushed more publicly for required safeguards and legislative involvement. Trump, however, has reiterated that overregulation could let China gain ground in the global AI race.

The White House's March legislative recommendations explicitly argued against establishing a new federal AI rulemaking body. Its preferred model is sector-specific oversight through current agencies plus industry-led standards. Yet the administration's June executive order illustrates a partial shift toward targeted national-security controls: it directs agencies to create classified cyber-capability benchmarks for advanced models and a voluntary pre-release access framework for models designated as "covered frontier models."

A Trillion-Dollar Table

The executives attending Tuesday's lunch collectively span the entire AI stack—from frontier model developers to cloud platforms, chip suppliers, and social networks. Their companies are pouring unprecedented sums into AI infrastructure.

Nvidia (NVDA), led by Jensen Huang, reported $96.2 billion in revenue in its latest quarter, including $89.0 billion from its Data Center segment, up 117% year over year. The chipmaker guided the following quarter to roughly $108 billion. Microsoft (MSFT)'s Azure cloud grew 43% in its fiscal fourth quarter, and the company projected $175 billion in calendar-2026 capital expenditures. Amazon (AMZN)'s AWS sales rose 37% to $42.2 billion in its latest quarter, and the company raised expected 2026 cash capex to $220 billion. Meta (META) reported approximately $31.1 billion in quarterly capex, largely on servers and data centers, with a stated 2026 outlook of $130–145 billion. AMD (AMD) posted record Q2 revenue of $11.5 billion, up 50%, with Data Center revenue more than doubling.

"We are in the middle of a generational buildout," said one executive who declined to be named. "The policy environment needs to keep pace with the investment."

Safety vs. Speed

The meeting comes amid a fierce debate over AI safety. Anthropic's Amodei has been vocal about the need for mandatory evaluation, monitoring, and potential legal safeguards before frontier capabilities become widely available. On the other side, many industry advocates argue that prescriptive rules could push investment and talent abroad.

The administration's stance is broadly pro-deployment: retain U.S. technological leadership, use existing agencies where possible, encourage industry standards, and intervene more narrowly when national-security risks are high. The March framework says Congress should avoid creating a new AI regulator.

But the June executive order signals a willingness to impose targeted controls. It directs agencies to develop classified benchmarks for advanced models and a voluntary pre-release access framework for "covered frontier models." The order does not define which models qualify, leaving key details to be determined.

State regulators and consumer advocates, meanwhile, are filling the void. In the absence of comprehensive federal legislation, states are advancing laws on discrimination, privacy, automated decision-making, and high-risk uses. This patchwork creates compliance challenges that the White House has sought to limit or preempt.

Global Stakes

Internationally, the AI race is increasingly a U.S.-China contest. Advanced models can affect cyber offense and defense, intelligence, military logistics, and semiconductor supply chains. The administration's principal argument against sweeping new restrictions is that they could cede advantage to China.

The same strategic framing was visible days earlier at the Trump–Xi dinner, which included an unusually large contingent of U.S. technology executives. That overlap underscores how closely AI business policy and U.S.-China diplomacy have become linked.

What to Watch

In the short term, market watchers will look for any public statement, voluntary safety commitments, or executive-action roadmap emerging from the lunch. They will also watch for clarity on which models qualify as "covered frontier models" and how voluntary pre-release testing will operate.

Longer term, the most plausible outcome is a hybrid U.S. regime: voluntary industry standards plus existing-agency enforcement, with tougher controls for national-security-sensitive frontier models. That would favor targeted standards over a single, powerful new AI regulator.

A White House spokesperson did not respond to a request for comment by press time. A spokesperson for Speaker Johnson's office declined to comment on the guest list.

Correction: An earlier version of this article misstated the timing of the Trump–Amodei dinner. It took place on September 27, not September 26.