AI Arms Race Explodes: $650B Capex Surge, Dow Hits 50K & OpenAI vs Anthropic Model Wars | February 7, 2026

Daily AI Blog

📋 Quick Takeaways

  • Big Tech commits a staggering $650 billion in combined 2026 AI capex — the largest corporate infrastructure buildout in modern history
  • Amazon shocks Wall Street with a $200B AI infrastructure plan ($50B above estimates), sending shares down 10%
  • Dow Jones crosses 50,000 for the first time ever, surging 1,206 points in a single session fueled by AI chipmaker rallies
  • NVIDIA CEO Jensen Huang declares the $660B AI capex buildout “sustainable” for 7–8 years, lifting NVDA shares 8%
  • OpenAI and Anthropic release rival flagship coding models within 15 minutes of each other in dramatic AI arms race escalation
  • Claude Opus 4.6 discovers 500+ zero-day vulnerabilities in open-source code during pre-release testing
  • Goldman Sachs reveals 6-month Anthropic partnership to automate banking operations with AI agents
  • $1 trillion wiped from software stocks as AI disruption fears reach fever pitch

💰 Big Tech AI Infrastructure & Market Shockwaves

Amazon Announces $200 Billion AI Infrastructure Spend — Stock Plunges 10%

Amazon revealed plans to invest $200 billion in AI and cloud infrastructure in 2026, smashing analyst expectations of approximately $150 billion by more than $50 billion. The announcement, made during Amazon’s Q4 2025 earnings report on February 5, sent shares tumbling as much as 10% on February 6, erasing roughly $240 billion in market capitalization.

CEO Andy Jassy defended the massive outlay during the earnings call, citing “very high demand” for AWS AI compute and describing AI as a “seminal opportunity.” AWS revenue grew 24% year-over-year to $35.6 billion — its fastest growth in 13 quarters. Amazon’s custom AI chips business reached $10 billion in annual revenue with triple-digit growth, while AWS backlog expanded 40%.

The company reported Q4 revenue of $213.4 billion (beating estimates) but EPS of $1.95 slightly missed the $1.98 consensus. Net income of $21.2 billion included $9.5 billion in pre-tax gains from Anthropic investments. Analysts at D.A. Davidson downgraded Amazon to neutral, warning that AWS is “scrambling to catch up through escalating investment.”

Sources: Bloomberg | CNBC | Motley Fool

Alphabet Plans Record $175–185 Billion AI Capex for 2026

Google parent Alphabet disclosed 2026 capital expenditure guidance of $175 billion to $185 billion — more than double its 2025 spend of $91.4 billion and far exceeding the Street estimate of $119.5 billion. The figure would represent more than Alphabet’s total capex over the past three years combined.

The company’s Q4 results were strong across the board: revenue climbed 18% to $113.8 billion, EPS rose to $2.82 (beating $2.63 estimates), and Google Cloud surged 48% year-over-year to $17.7 billion — the fastest growth rate among major cloud providers. Cloud backlog doubled year-over-year, reaching $240 billion.

CEO Sundar Pichai revealed that Gemini AI now has 750 million monthly active users, up from 650 million the prior quarter. He noted that “compute capacity” is what keeps him up at night, and said Alphabet will remain supply-constrained throughout 2026 despite the record infrastructure investment. CFO Anat Ashkenazi confirmed that approximately 60% of capex goes to servers and 40% to data centers and networking equipment.

Sources: CNBC | Bloomberg | Yahoo Finance | Fortune

NVIDIA CEO Jensen Huang: $660B AI Capex Buildout Is “Sustainable” for 7–8 Years

Speaking on CNBC’s Halftime Report on February 6, NVIDIA CEO Jensen Huang delivered a forceful defense of the AI infrastructure spending wave, declaring the combined $660 billion hyperscaler capex buildout “justified, appropriate and sustainable.”

Huang predicted the AI infrastructure expansion will continue for seven to eight years, calling it “the largest infrastructure buildout in human history.” He dismissed comparisons to the dot-com era, noting that unlike earlier tech booms, there is no idle infrastructure today — every GPU Nvidia has ever sold, including six-year-old A100 chips, is currently being rented.

Huang specifically praised both OpenAI and Anthropic as profitable ventures, stating: “If they could have twice as much compute, the revenues would go up four times as much.” NVIDIA shares surged nearly 8% on the day, adding over $300 billion in market capitalization and helping propel the Dow Jones past 50,000 for the first time.

Sources: CNBC | Bloomberg | Investing.com

Dow Jones Crosses 50,000 for the First Time in History

The Dow Jones Industrial Average surged 1,206 points (+2.5%) on February 6 to close at 50,115.67, breaking the 50,000 barrier for the first time in its storied 130-year history. The S&P 500 rallied 2% for its best day since May, while the Nasdaq Composite leaped 2.2%.

Chip companies led the broad-based recovery: NVIDIA jumped 7.8%, Broadcom climbed 7.1%, and semiconductor stocks across the board rallied. The rebound came after a brutal week of selling triggered by AI disruption fears and massive capex announcements. The rally was fueled by strong University of Michigan consumer sentiment data (highest since August), easing inflation expectations, and renewed confidence in AI infrastructure spending after Jensen Huang’s public defense of the buildout.

Traders at the New York Stock Exchange donned “DOW 50,000” caps as the milestone was achieved, marking the end of a 21-month journey from the 40,000 level.

Sources: AP/Fortune | ABC News | 24/7 Wall St.

🤖 AI Model Wars: OpenAI vs Anthropic

OpenAI Launches GPT-5.3-Codex — First Model Rated “High” Cybersecurity Risk

OpenAI released GPT-5.3-Codex on February 5, its most capable agentic coding model, claiming state-of-the-art performance on SWE-Bench Pro and Terminal-Bench 2.0. The model runs 25% faster than its predecessor and was notably used internally to help debug its own training and deployment — a milestone OpenAI described as AI systems “contributing to their own development.”

Critically, GPT-5.3-Codex is the first OpenAI model classified as “High capability” for cybersecurity under the company’s Preparedness Framework. CEO Sam Altman acknowledged this represents a new threshold where AI could “meaningfully enable real-world cyber harm, especially if automated or used at scale.”

As a result, OpenAI is deploying unusually tight controls: API access has been delayed, a “Trusted Access for Cyber” pilot program has been launched for vetted security professionals, and the company committed $10 million in API credits for cybersecurity defense research. The model is available to paid ChatGPT users but broader developer access remains gated pending safety reviews.

Sources: Fortune | OpenAI Blog | TechCrunch | The New Stack

Anthropic Launches Claude Opus 4.6 with 1M Token Context & Agent Teams

Anthropic released Claude Opus 4.6 on February 5, introducing a 1-million-token context window (in beta), dramatically improved document and financial analysis capabilities, and a groundbreaking new Agent Teams feature in Claude Code that enables multiple AI agents to collaborate in parallel on complex projects.

The model now tops the Finance Agent benchmark for financial analyst tasks and demonstrates a transformative leap in long-context performance — scoring 76% on MRCR v2 compared to just 18.5% for Sonnet 4.5, effectively eliminating the “context rot” problem where extended conversations degrade model output.

Anthropic’s head of enterprise product, Scott White, described the evolution as transitioning from “vibe coding” to “vibe working” — signaling AI’s expansion from software development into broad professional workflows. The release deepens integrations with Microsoft Office, including Excel interpretation and PowerPoint generation capabilities.

Sources: CNBC | Bloomberg | Axios

OpenAI vs Anthropic: Same-Day Model Launch in 15-Minute Showdown

In what may be the most dramatic moment yet in the AI arms race, OpenAI and Anthropic released their flagship coding models within 15 minutes of each other on February 5. Both companies had originally planned identical 10:00 AM PST launch times, but Anthropic moved its Claude Opus 4.6 release forward by 15 minutes, narrowly beating OpenAI’s GPT-5.3-Codex to market.

Both models represent a fundamental shift from code completion to semi-autonomous agentic AI capable of managing complex, multi-step professional workflows. On benchmarks, GPT-5.3-Codex led on Terminal-Bench 2.0, while Claude Opus 4.6 outperformed on reasoning and information synthesis tests like GDPval-AA and BrowseComp. The rivalry highlights diverging philosophies: OpenAI is prioritizing raw automation capabilities, while Anthropic emphasizes deep analytical collaboration with stronger safety guardrails.

Sources: TechCrunch | TradersUnion

Claude Opus 4.6 Discovers 500+ Zero-Day Vulnerabilities in Open-Source Code

Before its public release, Anthropic’s frontier red team tested Opus 4.6 in a sandboxed environment equipped with standard Python and vulnerability analysis tools — but no specialized instructions or training. Using only its out-of-the-box capabilities, the model discovered over 500 previously unknown zero-day vulnerabilities in open-source software, each validated by Anthropic staff or external security researchers.

In one remarkable case involving GhostScript, Claude turned to the project’s Git commit history after both fuzzing and manual analysis failed — demonstrating novel reasoning approaches to security research. In the CGIF case, Claude proactively wrote its own proof-of-concept exploit to confirm a vulnerability was real.

Logan Graham, head of Anthropic’s frontier red team, told Axios: “I wouldn’t be surprised if this was one of — or the main way — in which open-source software moving forward was secured.” Anthropic is now exploring ways to bring these vulnerability detection capabilities to the broader cybersecurity community.

Sources: Axios

Claude Sonnet 5 “Fennec” Leak Signals Imminent Launch

Developer forums and Google Vertex AI logs have revealed a model identifier claude-sonnet-5@20260203 with the internal codename “Fennec.” A 403 Forbidden error on the Vertex model ID (versus a 404 for nonexistent models) strongly suggests the model is built and awaiting activation within Anthropic’s infrastructure.

Early community testing of a non-thinking Sonnet 5 variant indicates coding performance surpassing Opus 4.5 at dramatically lower cost — potentially ~50% cheaper ($3/$15 per million tokens vs. Opus 4.5’s $15/$75). Strong math capabilities and a possible 1-million-token context window have also been discussed, though final specifications remain unconfirmed. Industry speculation points to a release during Super Bowl week or shortly thereafter.

Sources: Dataconomy | Technobezz

🏦 Enterprise AI Adoption

Goldman Sachs Partners with Anthropic to Automate Banking Operations

Goldman Sachs revealed on February 6 that it has spent the past six months with embedded Anthropic engineers co-developing AI agents to automate core banking functions, including trade and transaction accounting, client due diligence, and onboarding — processes traditionally resistant to automation due to their complexity and regulatory requirements.

CIO Marco Argenti described Claude as a “digital co-worker for many of the professions within the firm that are scaled, complex and very process intensive.” He said the bank was “surprised” at Claude’s capabilities beyond coding, noting its strength in parsing large datasets while simultaneously applying rules and judgment — skills critical for accounting and compliance work.

The partnership originated from Goldman’s testing of AI coding assistant Devin, which led to the discovery that Claude’s reasoning abilities transferred effectively to non-coding domains. Goldman CEO David Solomon had previously announced plans to “constrain headcount growth” as AI agents take on more operational roles. Goldman shares jumped 4.3% on the news, contributing to the Dow’s historic 50,000 milestone.

Sources: CNBC (exclusive) | Reuters/US News | PYMNTS

📉 Market Disruption & Software Selloff

$1 Trillion+ Wiped from Software Stocks as AI Disruption Fears Intensify

The iShares Software ETF lost nearly $1 trillion in market value over the past seven days, while six of the largest tech companies — Microsoft, NVIDIA, Oracle, Meta, Amazon, and Alphabet — collectively shed $1.35 trillion in market capitalization through Thursday’s close.

The selloff was triggered not by fears of an AI bubble, but by the opposite concern: that AI is on the verge of replacing traditional enterprise software. Anthropic’s Claude Cowork plugins for legal, finance, and marketing workflows were the initial catalyst, with subsequent model releases from both Anthropic and OpenAI deepening the panic.

Individual stock carnage has been severe: HubSpot is down 39% YTD, Figma has plunged 40%, Atlassian dropped 35%, and Shopify fell 29%. Thomson Reuters, LegalZoom, and RELX each declined over 15% in a single session. The WisdomTree Cloud Computing Fund has plummeted roughly 20% in 2026.

As KeyBanc analyst Jackson Ader warned: “While today it’s legal tech, tomorrow it might be sales or marketing or finance.” Bloomberg characterized the rout as unprecedented, noting: “Nothing quite rivals the rout rippling through stock and credit markets this week.”

Sources: Bloomberg | CNBC | ABC News | Reuters/Yahoo Finance

📊 Market Impact Analysis

The events of February 5–7, 2026 represent a watershed moment for the global technology landscape. Several critical dynamics are converging simultaneously:

The Capex Singularity: The combined $650 billion in AI infrastructure spending from just four companies exceeds the GDP of many nations. This capital concentration raises fundamental questions about returns on investment, energy consumption, and whether this represents sustainable growth or speculative excess.

The Model Arms Race Escalates: OpenAI and Anthropic releasing rival models within minutes of each other — with both systems now demonstrating potential cybersecurity risks — signals that competition is accelerating faster than safety frameworks can keep pace. GPT-5.3-Codex’s “High” cybersecurity rating is a first for the industry.

Software’s Existential Moment: The trillion-dollar software selloff represents the market’s real-time pricing of AI disruption risk. Unlike previous AI panics, this one is driven by actual product demonstrations rather than speculative fears — making it potentially more durable.

Finance Embraces AI Agents: Goldman Sachs deploying Claude for trade accounting and compliance represents a bellwether for the financial industry. If AI can handle the complexity and regulatory scrutiny of investment banking, few white-collar domains remain immune.

The Infrastructure Paradox: NVIDIA shares rallying 8% while Amazon dropped 6% on the same day illustrates a market bifurcation: investors are rewarding infrastructure providers while punishing the companies paying for that infrastructure — a tension that cannot persist indefinitely.

🔮 Looking Ahead

Key Trends to Watch:

  • Claude Sonnet 5 “Fennec” official launch — potentially the most cost-effective frontier model yet
  • GPT-5.3-Codex API access rollout and cybersecurity safeguard effectiveness
  • Software stock recovery or continued decline as AI agent capabilities expand
  • Q1 2026 hyperscaler earnings for early signals on AI capex ROI
  • Enterprise AI agent deployment patterns following Goldman Sachs’ lead
  • International AI Safety Report 2026 policy recommendations and government responses

Stay Updated: Follow us for comprehensive daily AI news, policy analysis, and investment tracking.

Last Updated: February 7, 2026, 8:00 PM CST