Nvidia to the rescue
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Thursday! Today, the media and political worlds are buzzing after ABC (owned by Disney) decided to take Jimmy Kimmel’s late-night show off-air. The reason? Kimmel said things about the murder of Charlie Kirk that made people mad. Welcome to the era of wrongthink. I’m not exaggerating; the FCC threatened Disney over the host’s comments.
POTUS has therefore neutered CBS via an aligned billionaire family’s financial activities, and now has ABC in the barrel. I presume that NBC will be next. And TikTok will likely wind up owned by some of the same money that defanged CBS, and backed the President. If you believed the current administration would be pro-free speech, I don’t know what to tell you. (More here from Axios). To work! — Alex
📈 Trending Up: AI-media deals (Meta edition) … AI in India … CEO dissent (behind closed doors) … Google in China … China’s HPC future … the Orbanization of America … Microsoft in Wisconsin …
📉 Trending Down: Privacy … benefits at Workday … public health … becoming a citizen … Live Nation and Ticketmaster …
Things That Matter
People Are Worried About AI: While in tech circles, AI is heralded as the next great revolution in human creativity and productivity, most folks see things differently. Per Pew, 50% of Americans are “concerned than excited about the increased use of AI in daily life,” a huge jump from 37% in 2021 and 38% in 2022.
53% of Americans think that AI will make us less able to think creatively
50% of Americans think that AI will make it harder for folks to “form meaningful relationships with other people”
It’s a closer split ont he question of AI helping (29%) or hindering our ability to solve problems (38%)
Why do we care? Negative public views concerning AI usage could hamper adoption; businesses that want to sell AI tools to the general public could face a steeper climb than they expect. I remain an AI bull, making this dataset a bit surprising in how negatively it describes the public view.
All Things IPO: StubHub’s IPO will not ignite another round of IPOs are mispriced and leave money on the table debate. After pricing at $23.50 per share for a valuation of around $8.6 billion, shares of the online ticketer fell 6.4% during its first day of trading.
Netskope will start trading today, after pricing at $19 per share, the upper-end of its IPO price range. RenCap pegs it as worth about $9 billion at that price. If the cybersecurity company also struggles in early trading, the market could decide that the summer’s IPO mania is behind us. Which would be a bother, as it’s been lovely having so many companies list.
Via, a transportation software company that listed earlier this month, has secured a partnership with Waymo to bring self-driving cars to public transit in an Arizona city. Call it an experiment, but I think having a city-based fleet of autonomous cars is going to be a community selling point in a few years’ time.
Nvidia Lends Intel A Hand: After the US government forced Intel into handing over around 10% of its equity for the disbursement of previously approved grants, another American entity is stepping up to the ‘buy some Intel’ plate. Nvidia will invest $5 billion into Intel shares at a price of $23.28 per share; Intel’s equity pumped 27% on the news of the deal, which includes planned work to bring Intel’s CPUs into Nvidia’s GPU products (for data centers), and Nvidia GPU chiplets into Intel’s personal computing offerings.
Investors are betting that Nvidia’s massive demand curve will help boost Intel’s fortunes. Nvidia is also in business with the government (via a forced 15% vig on sales to China), and is also now a co-shareholder with Uncle Sam in Intel. In short, welcome to state capitalism.
xAI’s Financial Projections: A recent string of executive departures from AI foundation model company xAI are a bit clearer today, thanks to WSJ reporting. Expected issues concerning management style (running a company at breakneck speeds under the tutelage of Elon Musk tends to churn through leadership) are being augmented by concerns about how the company’s financial situation is being handled:
Some xAI executives said they left because they were concerned that some of the company’s financial projections were unrealistic, the people said. They also raised questions internally about the role Musk’s family office, Excession, played in managing some of xAI’s cash and accounting.
I can help untangle that. What you need to know about xAI is that it is spending a lot of money very quickly to expand its compute footprint. Its first major compute cluster (Coloussus 1) was built at an incredible clip. With around 100,000 GPUs to start and another 100,000 added, xAI spent billions on the supercomputer. And to good effect, in terms of product. xAI’s Grok 4 model was generally well-received, and its Grok-Code-Fast-1 model has been a hit.
As we noted, Grok-Code-Fast-1 is priced so competitively that huge usage doesn’t add up to the sort of revenue scale we might expect.
So, what’s the issue? xAI is spending much more on its second supercomputer, called Colossus 2. Its first phase should feature more than 500,000 GPUs before rising to as many as a million or more. It’s super cool to have a supercomputer, no doubt. But they cost a mint. Colossus 1 was reported to have a $7 billion hardware cost, and Colossus 2 will cost double-digit billions.
Which is tricky for xAI, because despite a quick ramp to relevance in the AI model game, it’s not nearly as large in revenue terms as OpenAI and Anthropic, which were operating at $12 billion and $5 billion run rates a few weeks back. xAI will likely need to raise more capital through equity and debt sales to continue expanding its compute footprint. To do that, it needs a clear plan to generate enormous revenues. Quickly. And therein lies, per reporting, the tension.
Hey Look, M&A
Closing today, a few pieces of M&A have touched down that warrant our attention.
Atlassian will spend $1B on DX: DX builds tools that help companies sort out where their development teams spend time, and where bottlenecks exist. The gist is that building software is expensive in human capital terms, so you want to get max bang for your top-dollar buck. Atlassian buying the company makes sense as its products (Jira et al) are probably the tools that DX customers are trying to better understand how their teams use.
The Utah-based company raised little external capital, making its exit to Atlassian for “approximately $1 billion in cash and restricted stock” is incredible.
DX won’t harm Atlassian’s profit guidance, with the company saying in an SEC filing that deal will not change its “previously issued fiscal year 2027 non-GAAP operating margin target.”
Atlassian bought The Browser Company recently, bringing its agentic- browser under its purview. I don’t think it’s difficult to see how DX and TBC fit into the larger Atlassian if your developers need a thing, we make that thing strategy.
Fun question: When does Atlassian buy PostHog? Anyone want to lay a bet?
CrowdStrike will buy Pangea for $260M: Cybersecurity shop CrowdStrike will cough up a little over a quarter-billion dollars for Pangea Cyber, it told investors.
Pangea Cyber is interesting in that it was built to secure enterprise AI usage. That means visibility (what models are being used, and where), prompt safety (preventing injection attacks), and so forth.
For CrowdStrike, the purchase brings LLM-focused, native AI security into its larger mix of products. (Recall that CrowdStrike is best known for its work in endpoint security).
Investors are cheering the news, sending shares of the acquirer up 10% today. That’s more than $10 billion in new value, as the company was worth north of $100 billion before the deal dropped. (An analyst upgrade likely helped, too.)
More, please.