Compute Boom Meets Profitability Test.

Mazen Al Sabbagh

— 5 min read
Share

TL;DR

  • Compute demand, validated — but markets tempered: TSMC's record Q2 revenue of NT$1.27 trillion (~$39.6B) increased 36%, yet its ADRs declined ~3%. Investor focus has shifted to margins and capex transparency, not merely top-line growth, ahead of Thursday's earnings call.
  • Albany freezes the buildout: New York's executive order pauses permits for hyperscale data centers (50 MW+) for up to a year. This introduces a political constraint to the compute supply crunch, complementing existing physical limitations.
  • Apple v. OpenAI widens: New Bloomberg reports suggest the trade-secret litigation could directly impede OpenAI's nascent iPhone-rival hardware device, days before its anticipated IPO filing, rather than just impacting recruiting.
  • Half-year funding record: Global startup funding reached $510B in H1 2026, surpassing all of 2025. OpenAI and Anthropic alone secured 43% of this capital, underscoring concentrated investment in frontier AI.
  • Critical convergence point: July 17 marks the anticipated launch of Gemini 3.5 Pro, coinciding with the Shanghai World AI Conference, where Xi Jinping is confirmed to attend in person.

Lead Story: Compute Boom Meets Profitability Test

The AI industry's primary bottleneck this summer is compute capacity, and TSMC's latest financial disclosure provides tangible evidence. The foundry reported record second-quarter revenue of NT$1.27 trillion (~$39.6 billion), a 36% year-over-year increase, exceeding market expectations (TechTimes, AI Magazine).

June alone contributed NT$442.7 billion (~$13.8 billion), marking a 67.9% increase and the highest monthly revenue in the company's four-decade history, notably in a period traditionally prone to deceleration (Briefs). Comprehensive results and forward guidance are slated for Thursday, July 16.

These figures underscore a persistent pattern of demand exceeding supply. Google is reportedly rationing Meta's access to its Gemini models due to insufficient capacity (CNBC), concurrently committing $920 million monthly to SpaceX—totaling approximately $30 billion through mid-2029—for 110,000 Nvidia GPUs designated as "bridge capacity" (TechCrunch).

Further evidence includes OpenAI's weekend usage-limit resets following GPT-5.6 launch traffic, Anthropic's extended Fable 5 access promotion (now through July 19), and SK Hynix's projection of 2027 as the memory industry's most constrained supply year. The collective indicators are unequivocal: the primary bottleneck has shifted from model sophistication to raw silicon capacity.

Yet, the market reaction was muted. TSMC's ADRs declined 2.9% to approximately $421 post-announcement (FX Leaders, GuruFocus), signaling that record AI-driven revenue has become the market's expectation, not a differentiator.

Goldman Sachs maintains its view of AI demand as a "multi-year structural growth engine" for the chipmaker. However, Thursday's earnings call will be a critical examination of margins and capital expenditure efficiency: the market's inquiry has moved beyond the validity of the boom to its sufficient profitability for current valuations.

In Other News

New York froze new hyperscale data center permits — a national first. New York's Governor Kathy Hochul today signed an executive order instituting a one-year moratorium on state environmental permits for new hyperscale data centers (50 megawatts or more). This policy allows regulators to develop a comprehensive statewide environmental framework addressing grid, water, and air impacts (Governor's office, The Hill). The move coincides with TSMC's revenue report confirming seemingly insatiable demand, highlighting that the compute buildout now faces political as well as physical limitations. Its potential to set a precedent for other states merits close observation.

Apple's lawsuit now threatens OpenAI's hardware device itself. Bloomberg reports the ongoing trade-secret litigation initiated by Apple could directly impact OpenAI's strategic foray into hardware, specifically its "io Products" device. An injunction concerning designs linked to former Apple personnel would directly impede the development of this iPhone competitor (Bloomberg). This escalation surfaces days before OpenAI's anticipated IPO filing, as its broader rivalry with xAI continues (Bloomberg). OpenAI asserts it holds "no interest" in competitors' proprietary information.

Global startup funding hit a record $510B in H1. Data from Crunchbase indicates global startup funding reached a record $510 billion in the first half of 2026, exceeding the $440 billion total for all of 2025. Over 70% of Q2 capital was directed to AI companies (Crunchbase News, SiliconANGLE). Notably, OpenAI and Anthropic alone accounted for $217 billion—43% of the aggregate sum. The exit market experienced its strongest performance since 2021, primarily driven by SpaceX's $75 billion IPO.

The AI wealth-fund debate is moving from pitch decks to polling. OpenAI's proposal to grant the U.S. government a 5% equity stake—valued at approximately $42.6 billion—via a sovereign wealth fund, mirroring Alaska's Permanent Fund model, is gaining traction. CEO Sam Altman has directly engaged figures like Trump, Lutnick, and Bessent on this concept (CNBC, Axios). Concurrently, new survey data reveals 69% of U.S. workers support mandating AI firms to contribute equity to a public fund (Build Fast with AI), suggesting public sentiment may exceed even OpenAI's proactive offer.

X / Social Pulse

Sam Altman initiated a fresh line of competitive commentary today, publicly critiquing Anthropic's new "There's Hope in Hard Questions" campaign. His remarks, "i thought this was satire, kept looking for the handle to be spelled c1audeai or something," followed by "Come for the best model, stay because we don't treat you with contempt," signal an intensifying public rivalry (Yahoo Finance). Concurrently, the protracted public exchange between Elon Musk and Altman regarding Apple's lawsuit has extended into its fourth day (Slashdot). The 69% approval rating for AI wealth funds now circulates as validation that AI redistribution is transitioning from a theoretical concept to a palpable political imperative.

One to Watch

Critical Day: July 17. This Friday marks the anticipated General Availability of Google's Gemini 3.5 Pro, following its architectural overhaul. While rumored specifications include a 2M-token context and a "Deep Think" reasoning layer, official model documentation or pricing remains undisclosed (TechTimes). Google is nonetheless preparing the market, having extended Gemini in Chrome to UK desktop users today (9to5Google). Coinciding with this expected launch, the World AI Conference will commence in Shanghai, featuring Xi Jinping's inaugural in-person attendance. TSMC's earnings call on July 16 will set the immediate market context.

Quick Hits

  • Patent eligibility scrutiny: The Senate Judiciary Committee today reviewed patent eligibility under Section 101, revealing markedly higher invalidation rates for AI patents (Legis1, IPWatchdog).
  • Anthropic Artifacts expansion: Anthropic introduced public sharing and multiplayer editing for its Artifacts feature, alongside direct creation from Slack via Claude Tag (TestingCatalog).
  • Gemini in Chrome UK rollout: Gemini's Chrome integration has expanded to UK desktop users, offering tab comparison and native Gmail/Calendar functionality (Engadget).
  • ByteDance's Seedream 5.0 Pro: ByteDance launched Seedream 5.0 Pro, featuring image decomposition into 10+ editable transparent-PNG layers and in-image text rendering in ~14 languages, signaling advanced creative-AI capabilities from Chinese labs (ByteDance Seed, TestingCatalog).
  • Unitree Robotics IPO approval: China's Unitree Robotics secured final approval for a ~$619 million Shanghai STAR Market IPO, positioning it as the first humanoid-robot stock (Caixin).

Today's market narrative is characterized by the convergence of unconstrained demand and escalating friction. TSMC's unprecedented revenue figures validate a seemingly insatiable appetite for silicon, yet New York's regulatory freeze and Wall Street's indifferent response signal that the AI buildout is increasingly governed by political mandates and profitability metrics, rather than purely technological momentum. Thursday's earnings call will provide the initial assessment of this evolving landscape.

Sources

Lock in. M. mazen@thorterminal.com

Read more