• A version of this story originally appeared in Business Insider’s weekly VC Insider newsletter.
• Sign up for BI’s VC Insider newsletter here.

Ben here. Up until recently, it was a great time to be the founder of a neolab, the lavishly funded, research-first AI startups founded by top talent who defected from the likes of OpenAI or Google DeepMind.

We reported last month that Jeff Dean, Google’s former chief scientist, is seeking a valuation as high as $50 billion for his new AI science startup. OpenAI cofounder Ilya Sutskever’s Safe Superintelligence is worth $32 billion. Another OpenAI alum, Mira Murati, is raising a round for Thinking Machines Lab at a valuation of around $40 billion.

But lately, in conversations with multiple VCs around Silicon Valley, there is a growing divide about whether neolabs can make the economics work with their multibillion-dollar price tags and formidable competition from frontier labs.

One VC at a top firm, who asked that I not use his name, told me neolabs are the most hotly debated type of investment right now.

“Some would say you shouldn’t invest in any, and others would say this is basically all you should be investing in because software’s dead, and you either have to own models or hardware, and even if you enter at a billion, the company could be worth a trillion,” he said.

When I recently asked Menlo Ventures partner Matt Murphy at Business Insider’s Growth Mode conference which sector he thought was most overhyped, he immediately cited neolabs, which was noteworthy, given that Menlo invested in some of them.

Murphy pointed to an example from his time as an early board advisor at Google, which he said did so well because it “was like a match thrown into a bonfire.”

“The opposite thing you want to do is start a match with a bonfire,” Murphy told me. “That’s the risk of what’s going on now with companies raising such large rounds.”

Another Menlo partner, Deedy Das, posted last month about the “difficult game” of making neolab math work.

Das seemed a bit taken aback by the reaction to his post and later texted me: “Like all startups, many will fail, some will work.”

Picking which ones will work is why VCs get paid the big bucks.

Neolabs are a power-law bet taken to an extreme, Radical Ventures’ Aaron Rosenberg told me.

Investors know most will likely burn enormous sums on compute and talent without ever becoming profitable. But the rare winner could become a foundational AI company worth hundreds of billions or more.

“We’re confident that the best neolabs will attract ongoing support,” Rosenberg said. “Talent continues to transition out of the incumbents and academia into founder roles, and we see no signs of that slowing.”

Like Radical, New Enterprise Associates wrote a bullish view in June. One of the authors, Andrew Schoen, told me his views have not changed. “There are neolabs I know of that are finalizing billion-plus-dollar rounds of equity financing, and those processes are moving very smoothly and quickly,” he said.

The Buzz

Katie here. Many of tech’s biggest names gathered at the Ritz-Carlton in San Francisco on Saturday, at a black-tie event in the name of Menlo Ventures’ 50th anniversary.

Hollywood titan turned venture capitalist Jeffrey Katzenberg of Dreamworks fame asked the audience to join a prayer “for the blockbuster IPO of Anthropic,” a joking reference to Menlo’s early investment in the company.

I also spotted Morgan Stanley’s storied technology banker Michael Grimes, who is undoubtedly holding his own AI IPO prayers. Other attendees included an array of public company and startup CEOs, including those from Snowflake, Zscaler, Legora, and OpenRouter.

As Menlo partner Venky Ganesan quipped, there was more AUM (assets under management) in the room than in many countries, given the attendees’ net worth and the investment capital they manage.

Menlo isn’t the first venture firm to host a splashy anniversary party; Index Ventures recently went big for its 30th anniversary.

A person at a rival venture shop expressed disappointment that their firm doesn’t host such celebrations.

Menlo’s night ended with concerts from two super confidential Grammy Award winners trying to rake in those AI bucks!

Around the Cap Table

• We reported this week that Fireworks AI is seeking fresh funding at a valuation of between $30 billion and $40 billion, a dramatic jump just months after its last blockbuster round was closed.
• “You Can See Everything” has everyone talking. Ben thought it was absolute genius, if a bit long. It will make this year’s Oscars race exciting. See it in theaters before the experience is lessened by spoilers.
• Our colleague Rya Jetha explored the excitement and pitfalls of investing in robotics startups. Like neolabs, robotics companies are commanding big valuations with little to show for it, yet.
• Foursquare’s $0 acquisition ended with no payout for employees or investors, according to founder Dennis Crowley.

Post of The Week

The VC Insider team

Ben Bergman, chief correspondent, in Los Angeles. Katie Roof, editor at large, in San Francisco. Pranav Dixit, senior technology editor, in San Francisco. Leena Rao, executive editor, in Los Angeles.

• A New Mexico judge ordered Meta to pay $25 billion over how it handled its users’ data.
• It could be the largest penalty against Meta in history.
• Meta recently agreed to an $18 billion settlement over the impact of its tech on kids.

A judge in New Mexico ordered Meta to pay $25 billion for misleading users about how it handled their data and for failing to prevent misinformation on Facebook.

The ruling could be the largest penalty in Meta’s history. Last month, Meta settled a child-safety lawsuit brought by a group of US states for $18 billion.

The New Mexico case focused on Meta’s handling of user records that Cambridge Analytica used for campaign purposes during the 2016 election. New Mexico alleged that Facebook misrepresented the extent of users’ control over their data on the app.

“Facebook’s wrongdoing—over the course of years and systematically broadcast to the public—goes to the heart of Facebook’s core business practices,” Judge Francis J. Mathew wrote in the ruling. “It relates to the way in which Facebook induces people onto the platform and ultimately keeps them there.”

In a statement, Meta’s chief legal officer, C.J. Mahoney, said the penalty “makes it all but impossible to do business in the state of New Mexico.”

“At its core, this lawsuit seeks to impose a tax on speech to enforce new rules for what people can say online,” Mahoney said. “We are confident we will prevail on appeal.”

A jury ruled against Meta last month, finding the company committed nearly 50 million violations of a New Mexico law prohibiting deceptive trade practices. Each violation carries a penalty of up to $5,000.

Some Meta executives, including CEO Mark Zuckerberg, were deposed in the case. New Mexico first sued the company in 2021, and the case went to trial last month.

In internal Meta communications presented at the trial, Zuckerberg appeared to obfuscate methods for users to hide their data.

“Give them a chance to change their privacy settings but obviously don’t encourage it,” Zuckerberg wrote in one internal message.

Meta’s next move remains unclear. The company didn’t immediately respond to a request for comment on Wednesday.

This is a breaking news story. It will be updated.

• Nvidia vice president Laura Fay has retired after a decade at the company.
• Fay grew Nvidia’s enterprise marketing team from roughly 60 to 600 employees.
• Nvidia brought in Google Cloud’s Alison Wagonfeld as its first chief marketing officer this year.

Laura Fay, a longtime Nvidia marketing executive who reported to CEO Jensen Huang, retired in July after roughly a decade at the AI chip giant.

Fay formerly served as Nvidia’s vice president of global enterprise marketing, overseeing campaigns, events, analyst relations, and partner marketing. She said her team grew from roughly 60 to 600 people by the time she left.

Fay was among 36 employees on an internal list obtained by Business Insider last year of company leaders reporting directly to Huang. Prior to joining Nvidia, she spent 22 years at Cisco, concluding her tenure there as VP of global marketing.

Fay said Huang recruited her to build Nvidia’s enterprise marketing organization as the company looked to expand beyond gaming.

“I want to change the world, and I want you to change it with me,” she recalled him telling her on their first call.

Fay said she began considering retirement around her 10th anniversary in January, though a call from Huang congratulating her and expressing hope that she’d stick around for another decade persuaded her to stay a little longer.

Nvidia declined to comment.

Fay’s retirement came as Nvidia also brought in new marketing leadership. In February, the company hired Google Cloud executive Alison Wagonfeld as its first chief marketing officer.

The Wall Street Journal reported at the time that Wagonfeld would oversee all marketing and communications teams and assume responsibilities previously divided among several leaders.

Those changes follow another longtime marketing departure. Greg Estes, Nvidia’s former vice president of corporate marketing and developer programs, retired in 2025 after roughly 15 years at the company.

Several other longtime Nvidians have retired this year, including sales chief Jay Puri, chief accounting officer Donald Robertson, and VP of hardware engineering Robert Huang. Some retirees told Business Insider that Nvidia’s exponential stock run helped precipitate their decisions.

Nvidia’s communications organization has also undergone a shake-up as scrutiny around the company has intensified, with roughly 15 employees departing since 2025 amid a reorganization under new leadership.

Have a tip? Contact this reporter via email at gweiss@businessinsider.com or Signal at @geoffweiss.25. Use a personal email address, a non-work WiFi network, and a non-work device; here’s our guide to sharing information securely.

• SpaceX is offering employees a 20% pay bump to volunteer to help build its data centers in Memphis.
• The company is looking for “leaders” who can “remove blockers”, according to emails seen by BI.
• Elon Musk said on Tuesday he often spends “four days a week” in Memphis working on SpaceX’s data center buildout.

SpaceX is offering employees a pay bump to temporarily move to Memphis and help build its data centers.

Elon Musk’s rocket company is giving employees a 20% pay increase if they volunteer to spend six weeks assisting its critical supercomputer buildout, according to an email seen by Business Insider.

“We now need leaders — current and emerging — to guide teams and remove blockers as we bring more supercomputers online,” read the email, which was sent in early October to all SpaceX employees.

It specified that volunteers did not need to be managers already, adding that construction experience is a plus and that “drive matters most.”

Employees who put themselves forward should expect to spend six weeks on-site in Memphis, Tennessee, or Southaven, Mississippi, where SpaceX is currently expanding its huge Colossus 1 and 2 data centers to meet surging demand for its compute business.

Volunteers “must be in good performance standing,” with more than one year of tenure preferred, the email said. Employees who volunteer will receive a 20% premium on their base salary or on their hourly rate of pay for the period of their assignment, and be eligible for “bonuses tied to program success.”

SpaceX did not respond to a request for comment.

The email described the reallocation of employees from other parts of SpaceX to assist the company’s data center buildout as a “surge,” a tactic employed by the company to divert employees and resources to mission-critical projects.

In 2022, CNBC reported that SpaceX offered employees pay bumps of between 10%-25% to temporarily relocate to Texas as it attempted to launch its Starship megarocket.

The company has been offering “surging” employees a 20% premium on their salary to move to Memphis since at least August, according to another email seen by Business Insider that called for “SpaceX experts and aces” to work for the next three months on the data center buildout. The email said the program is intended to end on December 31, 2026, but could be extended.

SpaceX will cover travel expenses and logistics, both emails said.

SpaceX’s compute cash cow

The supercomputers in Tennessee and Mississippi, which were originally built by xAI before the startup was folded into SpaceX in February, have become a critical part of the company’s business.

In SpaceX’s earnings on Tuesday night, the rocket giant said AI revenue that includes leasing out its compute to other companies — including Anthropic and Google — had soared to more than $7 billion, nearly half SpaceX’s entire quarterly revenue.

On the analyst call, CEO Elon Musk said that SpaceX was being “hounded relentlessly” by companies looking to rent out its compute, adding that they could have rented it out several times over.

SpaceX has been shifting executives in addition to rank-and-file employees to oversee its aggressive expansion. Business Insider previously reported that the company had placed a former Starship executive in charge of building out its compute capacity in Memphis.

Musk told investors on Tuesday he was also spending most of his time in the city as he looked to push SpaceX toward a target of bringing an extra 10 gigawatts of compute online by the end of 2027.

“I’ve been spending, personally, a lot of my time in Memphis, often four days a week,” the SpaceX CEO said. He previously said he was living in an Airstream trailer on the site.

Do you work at SpaceX and have any information to share? Get in touch with this reporter at tcarter.41 on Signal or tcarter@businessinsider.com. Use a personal email address and a nonwork device; here’s our guide to sharing information securely.

• The internet is ablaze after cybersecurity researchers using AI found a bug in the PHP programming language.
• Greg Brockman said bad actors could use the exploit to steal private user data.
• The FBI and NSA have advised people to either take down websites or update their software.

OpenAI said that 70% of the world’s websites could be in peril after cybersecurity researchers using its AI models found exploits in a popular programming language.

The National Security Administration’s cyber division advised all web providers that use PHP, a widely used open-source language, to take their systems offline or update their software after OpenAI said one of its models found two “critical” flaws in the language’s infrastructure.

Greg Brockman, the president of OpenAI, said in an X post that hackers could exploit the vulnerability to steal private user data or modify websites.

“We consider this one of the most impactful security issues in recent memory,” OpenAI said in a blog post.

While OpenAI said the vulnerabilities appear to have been identified before any attackers discovered them, the public disclosure set off a race for companies and governments to update their websites and apply fixes before hackers could exploit the bugs.

“We are seeing early attacks emerging and are updating mitigations as new vectors show up,” Dane Knecht, Cloudflare’s chief technology officer, said in a post on X.

Numerous websites, including those belonging to governments and the media, went offline on Wednesday to address the issue.

PHP is a popular programming language that often handles the behind-the-scenes work powering websites that require user logins and online orders.

The critical flaw was first identified by Calif, a cybersecurity research company using GPT-5.6 Sol. Wiz, another cybersecurity company, independently uncovered the issues, OpenAI said. The companies are part of an OpenAI-led initiative called “Patch the Planet,” which aims to make the world’s digital infrastructure more secure.

OpenAI has also warned that as cybersecurity researchers continue to probe software vulnerabilities using its ever-advancing frontier models, more security failures will emerge.

An engineer at Microsoft pointed out that the PHP vulnerability was discovered with an older model.

“I don’t think the world is prepared for the sheer number of vulnerabilities newer models will uncover,” the engineer, Nicolas Bustamante, said on X.

The public disclosure on Wednesday by OpenAI and others of the existence of the exploits triggered urgent messages from leaders worldwide, including the Japanese Prime Minister.

“We have an urgent request for everyone to take action tonight,” said Sanae Takaichi.

• AWS expected to get $5 billion of Nvidia GB300 racks, followed by $3 billion more each month.
• Some of these GPU systems are being stored elsewhere due to lack of data center power.
• Amazon CEO Andy Jassy sees AI as a major opportunity, but getting capacity online is a challenge.

Amazon is racing to secure AI hardware earlier than before, even as data center power shortages leave some GPU systems waiting in warehouses.

Under a new Early Rack Receipt process, AWS expected to receive roughly $5 billion in Nvidia’s GB300 server racks at US facilities before final deployment, according to August internal emails reviewed by Business Insider. The company projected roughly $3 billion more in GB300 racks each month “for the foreseeable future,” one of the emails said.

Some AWS sites had room for more racks but not enough available power to turn them on, people familiar with the effort said, leaving finished equipment to be staged at intermediary warehouses until the facilities were ready.

At least one AWS team did not have “a clear forecasted path” for getting the incoming racks into data centers at the time, according to one of the emails.

Dark GPUs

A lack of new electricity capacity is one of several bottlenecks slowing the buildout of some AI data centers. This has led to a phenomenon known as dark GPUs, where AI chips and servers are purchased but can sit idle due to power supply issues or other challenges. (Dark GPUs can also refer to system design flaws that mean this gear isn’t fully used after being plugged in).

This is partly why financial pressure is mounting on big data center projects. Money must be spent upfront to secure supplies of Nvidia GPUs and other valuable AI components. But if some data centers aren’t ready to run, those early investments can tie up huge amounts of cash. Debt financing can fill in the gaps, but the market has become less willing to lend against some of these projects lately.

Trade-off

Amazon’s early-rack program is one piece of the cloud leader’s broader AI buildout, which is putting new pressure on its cash flow, as Business Insider exclusively reported this week. Amazon expects roughly $220 billion in capital expenditures this year — more than any Big Tech peer. Its free-cash-flow turned negative last quarter for the first time since 2023, and could stay negative for a while.

For AWS, the trade-off is securing scarce GPU racks before competitors do, even if some data centers are not yet ready to turn them on. But taking ownership earlier can tie up billions in equipment before it starts generating cloud revenue.

Previously, AWS took ownership of a finished rack only upon delivery to a data center, according to internal documents. Under the new approach, it can pay for and record certain racks earlier in the supply chain.

An Amazon spokesperson said the company routinely stages finished hardware near deployment sites so it can install it once capacity becomes available, calling the practice “best business practice.” The spokesperson added that Amazon has long-established systems for tracking infrastructure assets and disputed that the broader company lacked a path for getting racks into data centers.

Amazon CEO Andy Jassy previously said the company has a “clear line of sight” to strong returns from its AI investments, calling the technology a “once in a lifetime opportunity.”

“We’ve been through this cycle with the first big AWS growth wave, and liked the results,” Jassy wrote in Amazon’s annual shareholder letter. “We expect to feel similarly about this next wave.”

Have a tip? Contact this reporter via email at ekim@businessinsider.com or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.