The White House Places New Restrictions on H1-Bs
- 01AI Revenue Metrics Are Being Systematically Gamed
- 02AI Compute Costs Are Hitting an Operational Ceiling
- 03Immigration Policy Creates New Talent Supply Risk for Tech
- 04Big Tech Is Actively Neutralizing Consumer AI Startups
- 05Government Is a Sleeper AI Infrastructure Buyer
1. Key Themes
AI Revenue Metrics Are Being Systematically Gamed
A growing chorus of founders, investors, and finance professionals is calling out deliberate inflation of ARR figures at AI startups — with the implicit backing of top-tier VCs seeking favorable PR coverage.
"The reason many AI startups are crushing revenue records is because they are using a dishonest metric. The biggest funds in the world are supporting this and misleading journalists for PR coverage." — Scott Stevenson, CEO, Spellbook
AI Compute Costs Are Hitting an Operational Ceiling
Enterprise adoption of AI coding tools is running into a hard budget constraint: the cost of AI compute may now rival or exceed the cost of the human labor it's meant to augment.
"Companies pushing employees to use more AI are starting to hit a cost wall, with Microsoft reportedly canceling most internal Claude Code licenses and Uber blowing through its 2026 AI coding-tools budget in just four months, underscoring a broader problem: for some teams, the cost of AI compute may already exceed the cost of the employees using it."
Immigration Policy Creates New Talent Supply Risk for Tech
The Trump administration's new requirement that most temporary visa holders leave the U.S. to apply for green cards via overseas consulates introduces material operational risk for tech companies dependent on H-1B workers.
"Most temporary visa holders seeking green cards will have to leave the U.S. and apply through American consulates overseas, a shift that could disrupt tech workers on H-1B visas and the companies that employ them by adding new delays and uncertainty to the permanent residency process."
Big Tech Is Actively Neutralizing Consumer AI Startups
The shutdown of Huxe illustrates a clear pattern: consumer AI startups in feature-level categories are being rapidly commoditized by platform giants with distribution advantages.
"Huxe, an AI audio-generation app founded by former NotebookLM developers, is shutting down after raising $4.6 million, becoming the latest consumer AI startup to struggle as podcast-generation tools are rapidly copied by larger companies including Spotify, Google, Adobe, Amazon, ElevenLabs, and Meta."
Government Is a Sleeper AI Infrastructure Buyer
The U.S. intelligence community has quietly approved a $9 billion classified AI chip procurement — a signal that sovereign and government demand for AI infrastructure is a meaningful and underappreciated market.
"The White House has approved a secret $9 billion request to help U.S. spy agencies acquire the cutting-edge chips needed to run the latest AI models on classified systems after chip shortages left agencies like the CIA and NSA struggling to fully test and deploy tools from companies including OpenAI and Anthropic."
2. Contrarian Perspectives
Google, Not OpenAI, May Win the AI Race
Against the consensus narrative of OpenAI's dominance, one NYT columnist argues Google's structural advantages — advertising monetization, 900M Gemini users, and deep OS/app integration — position it as the likely long-term AI winner.
"New York Times columnist Brian X. Chen argues that the search giant is starting to win the AI race, citing Gemini's rapid growth to 900 million regular users, Google's ability to make money from AI through advertising, and the company's plan to weave Gemini into Search, Gmail, Docs, Maps, Android, and eventually Apple's Siri."
Voluntary AI Oversight Could Become De Facto Regulation
Top tech leaders — including Sacks, Musk, and Zuckerberg — successfully argued that even a voluntary government preview process for frontier AI models carries serious competitive risk, a non-obvious claim that reshaped White House policy.
"David Sacks, Elon Musk, and Mark Zuckerberg spoke directly with President Trump in order to derail a planned AI executive order by warning that even a voluntary system for previewing frontier models with the government could become a de facto approval regime that might slow U.S. AI development."
ARR Inflation Is Institutionally Enabled, Not Just Founder Opportunism
The more alarming framing isn't that individual startups are gaming metrics — it's that the largest VC funds are actively complicit, making this a structural integrity problem in the private markets.
"The biggest funds in the world are supporting this and misleading journalists for PR coverage." — Scott Stevenson, Spellbook
3. Companies Identified
| Company | Description | Why Mentioned | Quote |
|---|---|---|---|
| Spellbook | Legal AI startup | CEO publicly exposed ARR inflation practices across the AI startup ecosystem | "Scott at Spellbook did a great job of highlighting some of what you might describe as bad behavior on the part of some companies." — Jack Newton, Clio |
| Clio | Legal tech startup | CEO validated Spellbook's claims and pointed to YC's Garry Tan for proper metrics guidance | "The post brought much-needed awareness to the topic." |
| Huxe | AI audio-generation app | Case study in consumer AI startup failure due to platform commoditization | "Becoming the latest consumer AI startup to struggle as podcast-generation tools are rapidly copied by larger companies." |
| Microsoft | Enterprise software/AI | Canceling internal Claude Code licenses; hit AI compute cost ceiling | "Microsoft reportedly canceling most internal Claude Code licenses." |
| Uber | Ride-sharing/tech | Burned through its entire 2026 AI coding-tools budget in four months | "Uber blowing through its 2026 AI coding-tools budget in just four months." |
| Search/AI | Simultaneously criticized for degraded AI search UX and praised as potential AI race winner | "Gemini's rapid growth to 900 million regular users." | |
| SpaceX | Aerospace | Expected $80B IPO using traditional Wall Street book-building process | "Could become the largest and most complicated stock-market debut ever." |
| Social platform | Shares down ~40% YTD; threatened by Meta's new Facebook Groups standalone app | "Reddit shares fell nearly 6% after Meta began testing Forum." | |
| Binance | Crypto exchange | Billions in crypto transactions tied to Iranian regime financing despite prior AML guilty plea | "Including activity as recently as this month, despite repeated internal red flags." |
| Kalshi / Polymarket | Prediction market platforms | Subject of congressional probe into potential insider trading on political/military events | "Seeking information on how the companies verify users, enforce geographic restrictions, and detect suspicious trades." |
| Farther | AI-powered wealth management | Raised $150M Series D at $1B+ valuation led by General Atlantic | "Combines portfolio management, financial planning, risk analysis, and AI-assisted advisory tools." |
| Fresha | Beauty/wellness booking marketplace | Raised $80M from KKR growth fund at $1B+ valuation | "An 11-year-old London company… raised an $80 million round." |
| SignalPlus | Crypto derivatives infrastructure | Raised $40M from HashKey for institutional crypto options and structured products | "Develops institutional infrastructure for crypto derivatives trading." |
| The Path | AI mental health/coaching | Tony Robbins co-founded; raised $14.3M seed from Prime Movers Lab | "Creates AI-guided therapy and coaching applications." |
| Oorja Bio | Biopharma – pulmonary fibrosis | Raised $30M Series A for peptide-based therapies targeting lung disease | "Developing peptide-based therapies for fibrotic and cardiopulmonary diseases." |
4. People Identified
| Person | Description | Why Mentioned | Quote |
|---|---|---|---|
| Scott Stevenson | Co-founder & CEO, Spellbook | Publicly called out AI ARR inflation; post went viral in startup community | "The reason many AI startups are crushing revenue records is because they are using a dishonest metric." |
| Jack Newton | Co-founder & CEO, Clio | Validated Stevenson's claims; pointed to need for proper revenue metric standards | "Scott at Spellbook did a great job of highlighting some of what you might describe as bad behavior." |
| Garry Tan | President, Y Combinator | Referenced for publishing guidance on proper revenue metrics for AI startups | Referenced in an "explanatory post…about proper revenue metrics." |
| Satya Nadella | CEO, Microsoft | Dismantling Microsoft's senior leadership structure; building flatter teams around Copilot | "Efforts to dismantle the company's long-standing senior leadership structure and replace it with smaller, flatter teams." |
| David Sacks | White House AI & Crypto Czar | Lobbied Trump directly to kill AI executive order | "Spoke directly with President Trump in order to derail a planned AI executive order." |
| Elon Musk | CEO, SpaceX/Tesla; advisor | Co-lobbied against AI oversight executive order | Same as above. |
| Mark Zuckerberg | CEO, Meta | Co-lobbied against AI oversight executive order | Same as above. |
| Chun Wang | Cryptocurrency billionaire | Set to lead SpaceX's first crewed Mars mission | "Reportedly set to lead SpaceX's first crewed Mars mission, a Starship flyby of the Moon and Mars." |
5. Operating Insights
Police Your Revenue Metrics Before Investors or the Market Does
AI founders and CFOs should proactively audit how they define and report ARR. Using inflated metrics may generate short-term PR wins but creates serious downstream credibility risk — especially as this issue now has mainstream attention from both the press and top investors.
"Certain aspects of ARR shenanigans have been the subject of multiple other news reports and social media posts." YC's Garry Tan has already published guidance on proper metrics, signaling the standards conversation is moving fast.
Model AI Compute as a Line-Item Cost, Not a Productivity Assumption
Operators need to treat AI tooling as a tracked budget item with measurable ROI, not a blanket productivity investment. The Uber example — burning an annual AI budget in four months — is a warning that unconstrained AI tool access without governance creates cost overruns.
"For some teams, the cost of AI compute may already exceed the cost of the employees using it."
Build Immigration Redundancy Into Hiring Plans
Tech companies relying on H-1B talent should scenario-plan around extended absences or processing delays caused by the new consular application requirement. Contingency hiring pipelines, remote-work policies, and legal support infrastructure are now strategic necessities.
"A shift that could disrupt tech workers on H-1B visas and the companies that employ them by adding new delays and uncertainty to the permanent residency process."
6. Overlooked Insights
Institutional Crypto Infrastructure Is Attracting Real Capital
While consumer crypto captures headlines, institutional-grade crypto derivatives infrastructure is quietly raising significant growth rounds. SignalPlus's $40M raise from HashKey for options trading, market-making, and structured products suggests that institutional crypto rails are a durable investment theme — separate from the speculation narrative.
"Develops institutional infrastructure for crypto derivatives trading, including options trading systems, automated market-making, and structured product services."
Road-Embedded Energy Generation Is an Emerging Hardware Bet
Austria-based REPS raised $23.6M for systems that convert vehicle braking force into electricity via road-mounted hydraulic and electromagnetic modules — a novel infrastructure play at the intersection of transportation and energy that received zero mainstream coverage but attracted a meaningful funding round.
"Builds road-mounted energy generation systems that convert the braking force of vehicles into electricity using hydraulic triggers and electromagnetic conversion modules."