$725 Billion AI Spend Splits Wall Street.

Mazen Al Sabbagh

— 7 min read
Share

TL;DR

  • Big Tech's AI Commitment: Four leading tech firms pledged a combined $725 billion for AI infrastructure in 2026, a 77% increase year-over-year. Market reactions diverged: Alphabet gained 7% on Cloud revenue surge, while Meta dropped 9% after raising capex forecasts. (Bloomberg, CNBC, Yahoo Finance)
  • Anthropic's Soaring Valuation: The AI startup is exploring a $50 billion capital raise at a valuation potentially exceeding $900 billion, driven by a $40 billion revenue run rate from Claude Code, positioning it to surpass OpenAI. (Bloomberg, TechCrunch)
  • White House's Mixed Signals on Mythos: The administration is drafting an executive action to utilize Anthropic's Mythos for federal agencies, while simultaneously opposing its expanded access to 70 private companies. The NSA is reportedly using Mythos to find software vulnerabilities. (Bloomberg, WSJ via Bloomberg)
  • Musk Trial Focus Narrowed: Judge Gonzalez Rogers ruled the OpenAI trial will not delve into "catastrophe and extinction" issues, limiting Musk's framing of the case. Cross-examination by OpenAI and Microsoft continues. (NYT Live, CNBC)
  • China's Huawei Chip Shift: Major Chinese tech entities, including ByteDance and Alibaba, are rapidly securing Huawei Ascend 950 AI chips after DeepSeek V4's launch on Huawei hardware, signaling a strategic move towards domestic semiconductor independence. (Reuters)

Lead Story: The $725 Billion AI Investment and Its Market Verdict

The Q1 earnings reports from four key tech giants confirmed the unprecedented scale of AI infrastructure investment, yet the market's reaction was anything but uniform. The underlying narrative has shifted from the necessity of expenditure to the imperative of return on investment.

Alphabet emerged as a clear leader, reporting $109.9 billion in Q1 revenue, a 22% increase, with EPS of $5.11 significantly surpassing expectations. Google Cloud was the standout, contributing $20 billion in revenue, up 63% year-over-year, driven by substantial enterprise AI adoption. CEO Sundar Pichai acknowledged being "compute constrained," indicating demand outstripping current capacity. Alphabet shares responded positively, climbing 7% in premarket trading.

Microsoft's revenue hit $82.9 billion, with Azure accelerating to 40% year-over-year growth. CEO Satya Nadella articulated a vision of an "agentic computing era." The company revised its fiscal 2026 capital expenditure forecast to $190 billion, substantially above analyst models, with CFO Amy Hood noting expected "capacity constrained" conditions throughout 2026.

Amazon's performance was robust, with Q1 revenue reaching $181.5 billion and EPS of $2.78, nearly doubling estimates. AWS grew 28% to $37.6 billion, marking its fastest pace in 15 quarters. The Bedrock platform processed more tokens this quarter than in all prior years combined, with customer spend surging 170% quarter-over-quarter. Amazon's internal chip development business now boasts a $20 billion annual run rate.

In contrast, Meta, despite beating revenue and EPS estimates, faced a 9% share price decline. The company raised its 2026 capital expenditure guidance to $125-145 billion, attributing the increase to rising memory prices and data center costs for future model training and inference. Unlike its peers, Meta's spending increase lacked a correlating "compute constrained" narrative, inviting closer scrutiny from investors.

Collectively, the four companies project approximately $725 billion in AI-related capital expenditure for 2026, a 77% increase from the previous year. Both Alphabet and Microsoft hinted at even higher spending in 2027. The market's Thursday reaction—with the Dow benefiting from AI-driven construction demand while the Nasdaq lagged due to Meta's performance—underscored a critical divergence. Investors are no longer questioning the AI buildout itself, but rather its demonstrable conversion into sustained revenue and profitability.

In Other News

Anthropic Eyes $900 Billion Valuation, Could Leapfrog OpenAI. Reports from Bloomberg and TechCrunch indicate Anthropic is evaluating unsolicited offers for a $50 billion funding round at an astounding $850-900 billion valuation. This trajectory would place it ahead of OpenAI’s $852 billion post-money valuation. The ascent is fueled by Anthropic's rapidly expanding annual revenue run rate, now approaching $40 billion, largely attributed to the adoption of Claude Code and Cowork. The board is expected to finalize a decision next month, marking a dramatic revaluation from its $380 billion figure just three months prior. (Bloomberg, TechCrunch, PYMNTS)

The White House Wants Mythos for Itself — and Only for Itself. The administration's stance on Anthropic's Mythos model became starkly contradictory. While drafting an executive order to onboard Mythos for federal agency use, circumventing its supply-chain risk designation, the White House simultaneously informed Anthropic of its opposition to expanding Mythos access to an additional 70 private firms. This opposition cites security concerns from a recent Discord breach and potential compute constraints that could degrade government access. Concurrently, Bloomberg revealed the NSA is actively testing Mythos to identify vulnerabilities in prominent software, including Microsoft products, reportedly impressed by its efficiency. This emerging policy signals a government intent to secure exclusive access to Mythos's advanced offensive cyber capabilities, limiting its broader commercial distribution. (Bloomberg — NSA, Bloomberg — WH opposition, Bloomberg — AI memo, The Next Web, CNN)

Musk's Safety Arguments Hit a Judicial Wall. On Day 4 of the Musk v. OpenAI trial, Judge Yvonne Gonzalez Rogers issued a significant pre-trial ruling, explicitly stating that the proceedings "are not going to get into issues of catastrophe and extinction." This decision effectively curtailed Musk's attempts to frame the contract dispute as a broader civilizational safety argument. The cross-examination by OpenAI and Microsoft lawyers has since intensified. Wednesday's session featured combative exchanges, with OpenAI attorney William Savitt challenging Grok's safety record, citing documented instances of racist or inappropriate content generation by xAI's model. Savitt's line of questioning suggested Musk's safety concerns appeared to escalate whenever he lacked control, leading to multiple judicial interventions to maintain decorum. (NYT Live, Axios, Reuters, CNBC)

FISA 702 Heads to the Senate — With a Poison Pill Attached. The House passed a three-year extension of Section 702 by a 235-191 vote, but attached a ban on Central Bank Digital Currency (CBDC). Senate Majority Leader John Thune has declared this CBDC provision "dead on arrival," jeopardizing the entire bill ahead of tonight's midnight expiration. The Senate faces the options of stripping the CBDC ban and returning the bill to the House, or passing a separate stopgap extension—both requiring potentially unavailable floor time. Opponents across both parties remain focused on the lack of a warrant requirement for queries involving Americans' data, particularly given the FBI's 34% increase in such searches in 2025, and the broader implications for AI-enabled surveillance. (NPR, Fox News, Nextgov, The Hill)

X / Social Pulse

The market's reaction to Anthropic's potential valuation surge dominated discourse on AI Twitter throughout Thursday morning. The prospect of an AI startup surpassing OpenAI's valuation, especially after a substantial raise just two months prior, ignited a polarized debate. Bulls pointed to the $40 billion annual recurring revenue (ARR) as definitive proof of Claude Code's transformative impact on Anthropic's financial trajectory. Conversely, bears questioned whether a $900 billion valuation for a company with such a high compute burn rate was realistically pricing in a decade of sustained dominance.

Concurrently, the Q1 earnings reports continued to fuel discussions regarding the implications of being "compute constrained." Sundar Pichai, Satya Nadella, and Amy Hood all utilized variations of this phrase, prompting analysis into its true meaning within the context of rapidly escalating demand. Meta's decision to increase capital expenditure without an accompanying "constrained" narrative, however, drew a distinct form of scrutiny regarding the efficiency and strategic rationale behind its investment.

Mid-morning, Bloomberg's report on the NSA's testing of Mythos generated sharp commentary. The apparent contradiction of the U.S. government simultaneously blacklisting Anthropic due to perceived risks while utilizing its most advanced model to probe Microsoft's vulnerabilities, and then blocking its broader distribution, highlighted a complex, and arguably hypocritical, approach to frontier AI governance.

One to Watch

The Huawei-DeepSeek Axis Is Reshaping China's AI Supply Chain. A Reuters exclusive confirms that major Chinese internet firms—ByteDance, Tencent, and Alibaba among them—are aggressively seeking Huawei Ascend 950 AI chips. This scramble follows DeepSeek V4’s recent launch, specifically optimized for Huawei hardware. This strategic pivot signals a decisive move toward domestic semiconductor independence.

The significance extends beyond mere procurement. The Ascend 950PR is China's only domestic chip capable of supporting the compressed numerical formats essential for efficient AI inference, outperforming Nvidia's export-restricted H20. While still trailing the H200, the 950PR's emergence, with mass production initiated in April and full-scale shipments slated for the second half of 2026, positions Huawei to deliver an estimated 750,000 units this year.

This development parallels the Western AI narrative: both sides face an AI demand that outstrips available compute, yet their supply chains are diverging. While American hyperscalers commit staggering sums to internal buildouts, Chinese tech giants are strategically consolidating around Huawei. The potential for DeepSeek's V4-Pro pricing to significantly decrease with scaled Huawei supernode shipments could introduce an order-of-magnitude cost advantage in inference, reshaping global market incentives.

Quick Hits

  • Anthropic's BioMysteryBench demonstrated Claude Mythos solving 30% of bioinformatics tasks beyond human expert capabilities, though Anthropic acknowledged the model's approach was "stumbling onto rather than reproducing" these solutions. (ResultSense)
  • Amazon's custom chip business achieved a $20 billion annual revenue run rate, exhibiting triple-digit year-over-year growth and establishing itself as one of the fastest-growing semiconductor operations globally. (About Amazon)
  • Q1 2026 global venture funding reached an all-time record of $300 billion, with investment in foundational AI startups during this single quarter doubling the total for all of 2025, per Crunchbase data. (Crunchbase)
  • Rogo secured $160 million in Series D funding, led by Kleiner Perkins, with participation from Sequoia, Thrive Capital, Khosla Ventures, and J.P. Morgan Growth Equity Partners, bringing total funding for the agentic AI platform for finance to over $300 million. (PR Newswire)
  • Connecticut's Senate passed an AI Bill of Rights (SB 5) by a 32-4 vote, instituting regulations for frontier AI model developers and establishing new protocols for youth social media and chatbot usage. (CT Mirror)

The earnings reports today solidified the thesis: AI demand is robust, accelerating, and consistently outstripping available supply across both Western and Eastern markets. Investors are now discerning, rewarding clear conversion of AI expenditure into tangible revenue growth, and scrutinizing spending that lacks a direct, quantifiable return. Yet, the day's more complex development might be Anthropic's position: simultaneously navigating a government blacklist and a secretive embrace by the NSA, its most critical model being utilized for national security while commercial distribution is restricted. Now, on the cusp of becoming the world's most valuable AI startup, its market incentives and geopolitical entanglements are unparalleled. As the Musk trial progresses, FISA faces a midnight deadline, and Huawei chips become a strategic battleground, the global AI landscape continues its rapid, multidirectional reordering, outpacing conventional institutional frameworks.

Sources

Lock in. M. mazen@thorterminal.com

Read more