The Profit Pivot

The tech industry's decade-long experiment with capital abundance is ending, and today's headlines reveal the accounting. After years where growth justified almost any burn rate, we're watching a synchronous shift across the stack: companies that once measured success in user acquisition are now being forced to prove they can generate actual cash.

The SpaceX IPO filing represents more than just a massive liquidity event. It's a signal that even the most capital-intensive, long-horizon businesses can no longer rely on private markets to indefinitely fund their ambitions. Meanwhile, Anthropic's path to profitability in just its second quarter reveals how quickly the economics of foundation models can flip once deployment reaches scale. This isn't just about cutting losses. It's about demonstrating that AI infrastructure companies can achieve viable unit economics faster than skeptics predicted.

Meta's layoffs to offset AI investments show the other side of this transition: established players are reallocating resources, not expanding headcount, to fund their next generation of products. The calculus has changed. Growth still matters, but the market now demands proof that growth eventually converts to profit. What we're witnessing isn't belt-tightening. It's recalibration toward sustainable business models across every layer of the industry.

Deep Dive

SpaceX's IPO filing exposes the new math of mega-scale ventures

SpaceX's S-1 filing reveals something more significant than its potential $1.75 trillion valuation. It shows that even the most ambitious, longest-horizon technology companies can no longer avoid the fundamental question: when does this become a sustainable business?

The numbers tell a sobering story. SpaceX generated $18.67 billion in revenue last year but lost $4.9 billion while spending $20.7 billion in capital expenditures, nearly double the prior year. Starlink alone brought in $11 billion, which means the core rocket business and newly merged xAI operations are still deeply unprofitable. The company projects a $28.5 trillion addressable market, with $22.7 trillion of that coming from enterprise AI applications that don't yet exist at meaningful scale.

For VCs and founders, this filing is instructive. SpaceX represents the best-case scenario for capital-intensive moonshots: legitimate technical breakthroughs, real revenue traction, and a founder with unmatched ability to command attention and capital. Yet even SpaceX must now answer to public markets about its path to profitability. The S-1's risk factors explicitly acknowledge that "several of our anticipated market opportunities do not currently exist" and that its "substantial level of indebtedness could materially adversely affect our financial condition."

The implications extend beyond SpaceX. If the most celebrated private company of the past decade needs public capital and faces questions about its burn rate, every other founder building long-horizon infrastructure businesses should expect similar scrutiny. The private markets that once funded these visions indefinitely are demanding exits. And exits mean accountability to investors who want growth narratives backed by credible paths to cash generation. SpaceX's willingness to go public despite its losses signals that the era of patient private capital has limits, even for the most ambitious builders.

Anthropic's profitability timeline resets expectations for foundation model economics

Anthropic reaching profitability in its second quarter would mark a watershed moment for AI infrastructure companies. Revenue more than doubling to $10.9 billion quarter-over-quarter demonstrates that foundation model businesses can achieve viable economics far faster than skeptics predicted.

The timing matters. OpenAI is reportedly preparing its own IPO, and the competitive dynamic between these two companies now includes a race to prove sustainable business models. Anthropic's momentum comes from Claude's growing adoption among professionals and deliberate moves to diversify its customer base beyond enterprise deals. The company has secured major compute contracts, including $1.25 billion monthly from Anthropic through May 2029, providing revenue visibility that investors value.

For founders building in the AI stack, Anthropic's trajectory suggests that foundation model businesses can reach profitability within years, not decades, if they achieve product-market fit and scale deployment efficiently. This is faster than many infrastructure businesses historically required. However, the Wall Street Journal's caveat that Anthropic may not remain profitable throughout the year due to compute costs highlights the precarious economics. Operating leverage exists, but maintaining it requires balancing growth investments against near-term profitability.

VCs evaluating AI infrastructure investments now have a clearer benchmark. The question shifts from whether foundation model companies can ever be profitable to how quickly they can get there and whether they can stay there while still investing in the next generation of capabilities. Anthropic's ability to cross into profitability while competitors still burn billions validates that the business model works at scale, but it also raises the bar. Startups that can't demonstrate a credible path to similar unit economics within a comparable timeframe will face harder questions about their strategic viability.

Meta's 8,000 layoffs signal resource reallocation, not retrenchment

Meta cutting 10 percent of its workforce isn't a retreat from AI. It's a wholesale reorganization to fund it. The company is simultaneously laying off 8,000 employees, reassigning 7,000 others to AI initiatives, and closing 6,000 open roles while nearly doubling capital expenditures to $115 billion to $135 billion this year for AI infrastructure and its Meta Superintelligence Labs.

This represents a different kind of layoff than the industry saw in 2022 and 2023. Those cuts followed pandemic over-hiring and collapsing digital advertising markets. These cuts are strategic, designed to redirect payroll dollars toward compute and research talent while maintaining overall investment levels in AI. The message to employees is clear: Meta isn't shrinking its ambitions, but it's making hard tradeoffs about where to deploy resources.

For tech workers, Meta's approach previews what resource allocation looks like in the profit-focused era. Companies are no longer expanding headcount to chase every opportunity. They're consolidating teams, cutting roles that don't directly support strategic priorities, and redirecting savings toward capital-intensive bets. The 7,000 employees being reassigned to AI work represent a forced march toward the company's future, whether those employees chose that path or not.

Founders should read Meta's move as a signal about capital allocation discipline. When established players with massive cash flows are cutting to fund their next platform, startups can't assume they'll have the luxury of patient capital to explore adjacent opportunities. Focus matters more than ever. The companies that survive this transition will be those that can articulate not just what they're building, but what they're explicitly choosing not to build in order to reach profitability faster.

Signal Shots

Nvidia bets $200 billion on AI agent CPUs : Nvidia's new Vera CPU has already generated $20 billion in sales this year, and CEO Jensen Huang projects a $200 billion addressable market for processors designed specifically for AI agents rather than traditional cloud workloads. The chips prioritize token processing speed over core count, optimized for agents running autonomous tasks. This positions Nvidia to own the infrastructure layer for agentic AI, not just model training. Watch whether hyperscalers build competing agent-specific silicon or accept Nvidia's architecture as the standard, and whether Vera's economics hold up as inference costs continue falling.

OpenAI races toward September IPO : OpenAI is preparing to file confidentially for an initial public offering that could happen as soon as September, working with Goldman Sachs and Morgan Stanley. The timing comes immediately after Elon Musk's lawsuit against the company failed, removing a major legal obstacle. Going public while still burning significant capital represents a bet that public markets will value growth trajectory over near-term profitability, contrasting with Anthropic's profitability push. Watch how OpenAI structures its governance to satisfy public market investors while maintaining its unusual corporate form, and whether it can sustain growth momentum through the IPO process.

Sam Altman offers every YC startup $2 million in OpenAI tokens : Altman proposed investing $2 million in compute credits into all 169 startups in Y Combinator's current batch in exchange for equity via uncapped SAFEs that convert at the Series A. The deal eliminates a major early expense for AI startups while locking them into OpenAI's platform and giving OpenAI equity upside. As inference costs fall, what OpenAI gives away today costs less to produce tomorrow, making the equity increasingly cheap. Watch whether Anthropic or other model providers counter with similar offers, and whether startups accepting the deal find themselves over-indexed on a single infrastructure provider when they need to optimize costs later.

First feature-length AI film costs $500,000 and two weeks : Higgsfield AI premiered Hell Grind at Cannes, a 95-minute fully AI-generated film that cost $500,000 to produce, with $400,000 spent on compute. Traditional indie features typically cost $1 million to $5 million and take months to produce. The economics suggest AI tools could democratize feature film production, though quality and artistic merit remain open questions. Watch whether traditional studios adopt AI to slash production budgets or resist to protect existing workflows, and whether film festivals continue accepting AI-generated entries as the technology improves and becomes more prevalent.

Nvidia doubles startup investment portfolio to $43 billion : Nvidia's stakes in private companies jumped from $22 billion to $43 billion in a single quarter, with $18.5 billion in new purchases, far outpacing the $649 million it invested the previous quarter. This doesn't include the $30 billion OpenAI commitment or public company investments. The acceleration suggests Nvidia is aggressively building an equity portfolio across the AI stack, positioning itself to profit from the entire ecosystem beyond chip sales. Watch whether this investment pace continues and if Nvidia starts taking board seats or more active roles in portfolio companies, potentially creating conflicts as it sells to their competitors.

Scanning the Wire

Manus co-founders seek $1B to buy back company from Meta : Beijing ordered Meta to unwind its $2 billion acquisition of the Chinese-founded company, forcing the founders into talks to raise over $1 billion for a buyback that highlights growing regulatory pressure on cross-border tech deals. (Bloomberg)

AMD commits $10B to Taiwan chip production : The company will invest over $10 billion in Taiwan's semiconductor industry for advanced AI chip packaging, with TSMC ramping up production of AMD's next-generation Venice processors to meet surging demand for AI infrastructure. (Wall Street Journal)

Truecaller launches eSIM business : The caller ID company is diversifying revenue streams by offering eSIM plans ranging from 1GB over seven days to 20GB over 30 days, initially available in 29 countries as it moves beyond its core identification service. (TechCrunch)

Tesla Full Self-Driving expands in Europe and China : After launching in the Netherlands and Lithuania, Tesla's driver assistance system is now available in China following years of regulatory delays, as the company accelerates international deployment while local EV rivals advance their autonomous capabilities. (TechCrunch)

GitHub breach exposes 3,800 internal repositories : A poisoned VS Code extension installed on an employee device gave attackers access to roughly 3,800 internal repositories, with threat group TeamPCP claiming responsibility and advertising the stolen code for sale starting at $50,000 as supply chain attacks intensify across developer tooling. (VentureBeat)

Intuit cuts 17 percent of workforce : The tax software maker is reducing headcount as investors worry that generative AI models could threaten traditional software businesses, signaling broader industry concern about AI disruption to established revenue models. (CNBC)

Climate tech pivots to critical minerals : Climate technology companies are reframing their value propositions around critical mineral production rather than pure decarbonization, adapting to weakened policy support in the second Trump administration while Boston Metal raises $75 million to produce metals beyond its original green steel focus. (MIT Technology Review)

Cohere releases Command A+ open model : The Canadian AI lab launched a sparse mixture-of-experts model with 218 billion total parameters and 25 billion active parameters, optimized for agentic tasks and released under Apache 2.0 license, its first fully open model as it competes with Meta and Mistral. (VentureBeat)

Outlier

Steel startup becomes metals hedge : Boston Metal raised $75 million to pivot from green steel production into critical minerals extraction, rebranding environmental technology as supply chain security. The shift reflects how quickly climate tech companies are abandoning decarbonization narratives in favor of geopolitical positioning. When cleantech becomes mineral tech within a single funding cycle, it signals that climate policy has lost its capacity to sustain entire industries. The future of environmental technology may depend less on carbon reduction mandates and more on whether it produces materials that governments consider strategically essential. Resource nationalism is quietly replacing sustainability as the organizing principle for industrial policy.

The strangest thing about this moment isn't that companies are finally chasing profit. It's that we spent a decade pretending they didn't have to, and now everyone's shocked the experiment ended. See you next time.