Dear reader, Welcome to our monthly AI digest! February 2026 marked a pivotal moment when AI companies stopped optimizing for benchmarks and started optimizing for businesses. From OpenAI introducing advertising to ChatGPT, to Anthropic's "vibe working" era with Claude Opus 4.6, to Google's reasoning breakthrough with Gemini 3.1 Pro—the industry shifted decisively from research labs to enterprise deployment. Here are the major stories that shaped the month.

On February 9th, OpenAI began testing advertisements in ChatGPT for US users on Free and Go subscription tiers, marking the most significant monetization shift since the chatbot's 2022 launch. The move came one day after rival Anthropic's Super Bowl commercial lampooned the idea of AI chatbot ads, triggering a public spat between CEOs.

Pricing Structure:

How Ads Work:

Privacy Commitments:

The Anthropic Feud: OpenAI CEO Sam Altman called Anthropic's Super Bowl ads "dishonest," accusing the company of being "elitist" for serving "an expensive product to rich people." Anthropic's commercial featured glassy-eyed actors playing AI chatbots delivering advice alongside poorly targeted ads, highlighting how advertising could disrupt user experience.

Altman framed the ads as necessary to "bring AI to billions of people who can't pay for subscriptions," positioning OpenAI as the democratizing force against Anthropic's premium-only model.

Industry Reaction: The advertising rollout represents a calculated risk. OpenAI COO Brad Lightcap described the approach as "iterative," emphasizing that trust and privacy are "non-negotiable." Critics fear ads could influence ChatGPT's responses despite OpenAI's claims that "ads do not influence the answers ChatGPT gives you."

With hundreds of millions of weekly users, ChatGPT represents a potential goldmine for advertisers. The question is whether the $60 CPM premium—justified by ChatGPT's high engagement and decision-making context—can sustain itself in a market where users have alternatives like ad-free Claude and Gemini.

On February 5th, Anthropic released Claude Opus 4.6, positioning it as the frontier model for "vibe working"—a term coined by Anthropic's Head of Product Scott White to describe how AI is moving beyond software development into general knowledge work.

Breakthrough Capabilities:

Adaptive Thinking: Opus 4.6 introduces contextual reasoning adjustment—the model picks up clues about how much to "think" based on task complexity, balancing intelligence, speed, and cost dynamically.

Enterprise Integration:

Market Impact: Following Opus 4.6's release and the broader Cowork plug-ins announcement earlier in the week, traditional software-as-a-service (SaaS) stocks experienced massive sell-offs. The WisdomTree Cloud Computing Fund fell over 20% year-to-date as investors panicked that AI agents would render legacy enterprise software obsolete.

Pricing Remains Competitive: $5 per million input tokens and $25 per million output tokens, with up to 90% savings through prompt caching—unchanged from Opus 4.5 but significantly undercutting GPT-5.2's pricing.

Enterprise Adoption Surge: According to Andreessen Horowitz's January 2026 survey, Anthropic's share of enterprise production deployments soared from near-zero in March 2024 to 44% by January 2026, second only to OpenAI's 77%.

Real-World Examples:

On February 19th, Google released Gemini 3.1 Pro—the company's first ".1" increment, signaling a targeted intelligence upgrade rather than broad feature expansion.

Performance Breakthrough:

Key Improvements:

Aggressive Pricing:

This pricing directly challenges Anthropic's $5/$25 per million tokens for Opus 4.6 and OpenAI's premium tiers.

Availability: Gemini 3.1 Pro rolled out to the Gemini app with higher limits for Pro/Ultra subscribers, NotebookLM (Pro/Ultra exclusive), and developer platforms including Vertex AI, Google AI Studio, Gemini CLI, Android Studio, and Gemini Enterprise.

Developer Feedback:

In the second week of February, four major tech companies—Google, Amazon, Meta, and Microsoft—announced a combined $650 billion in AI infrastructure investments for 2026, representing a staggering 67% increase from 2025's $381 billion.

Individual Commitments:

Market Reaction: Despite the historic investment announcement, the four companies' combined market capitalization decreased by more than $950 billion. Investors expressed skepticism that returns would be commensurate with the massive capital deployment, questioning whether AI revenue growth could justify the spending.

OpenAI-Cerebras Deal (February): OpenAI agreed to purchase up to 750 megawatts of computing power over three years from Cerebras Systems in a multi-billion-dollar contract valued at over $10 billion. The deal deploys Cerebras' wafer-scale AI chips to accelerate ChatGPT inference and scaling, reducing reliance on Nvidia while diversifying beyond Microsoft Azure.

Micron's $100B Megafab (February): Micron Technology broke ground on a $100 billion semiconductor megafab in Onondaga County, New York—the largest U.S. manufacturing facility of its kind and the biggest private investment in state history. The campus will house up to four fabrication plants producing DRAM and high-bandwidth memory (HBM) critical for AI workloads, generating around 50,000 jobs over 20+ years.

TSMC US Expansion (February): Taiwan Semiconductor Manufacturing Company posted record profitability driven by surging AI chip demand and doubled down on US expansion with additional Arizona fabs, bringing total investment to $165 billion. Management guidance projects robust 2026 revenue growth with $52-56 billion capex (up 27-37% YoY), 60-80% allocated to advanced processes.

Implications for Manufacturing: Analysts predict this infrastructure race will have two major impacts:

On February 27th, Defense Secretary Pete Hegseth declared Anthropic a "supply chain risk," and President Trump directed every federal agency to stop using Claude technology, with six months to phase it out.

Background: Anthropic's usage policy prohibits directly using Claude for domestic surveillance or lethal autonomous weapons. These restrictions led to FBI and Secret Service agents being unable to use the tool and created tensions with the Pentagon.

According to Financial Times reporting in mid-February, Hegseth threatened to cut Anthropic out of the DoD's supply chain or invoke the Defense Production Act to assert unrestricted use without an agreement.

Anthropic's Response: The company announced it would challenge the supply chain risk designation in court, maintaining that its safety restrictions align with responsible AI development.

The Irony: Despite the ban, Claude was reportedly used by the military during US strikes on Iran, highlighting the gap between policy and operational reality.

At the REAIM summit, major countries including the United States and China refused to sign a joint declaration on responsible AI deployment in the military domain, signaling continued divergence in AI governance approaches.

Chile's National Centre for Artificial Intelligence (Cenia) launched Latam-GPT, an open-source AI language model built specifically for Latin America to combat bias in US-dominated AI systems.

Key Details:

The initiative represents a significant step toward AI sovereignty, as regions seek to avoid "Silicon Valley AI colonization" of their cultures.

Perplexity Model Council (February 7): Revolutionary system running multiple frontier AI models (Claude, GPT-5.2, Gemini) in parallel to generate unified, cross-validated answers. The approach presents results from different models while noting commonalities and differences, significantly improving reasoning quality and reducing hallucination errors.

Meta Manus AI Integration (February): Meta embedded Manus AI—its autonomous agent technology acquired in late 2025—directly into Ads Manager workflows. The placement integrates agent capabilities into daily advertiser workflows for multistep tasks like market research, report building, and campaign analysis.

DBS Bank AI Payments Pilot (February): Singapore's DBS became the first Asia Pacific issuer to pilot Visa Intelligent Commerce (VIC) for everyday payments, successfully completing transactions using "AI-ready credentials"—tokenized card details that AI agents can use to make purchases with advanced authentication and intent-driven controls.

GPT-5.3-Codex-Spark (February 13): OpenAI released a smaller ultra-fast coding model optimized for real-time use in Codex, capable of generating over 1,000 tokens per second on low-latency hardware.

Claude Sonnet 4.6 (February 17): Anthropic released a mid-tier model priced the same as Sonnet 4.5, maintaining the balance between capability and cost.

NASA Mars Rover Planning (February 2): NASA's JPL successfully used Claude AI to plan a 450-meter path for the Perseverance rover on Mars, analyzing years of rover data and writing commands in Rover Markup Language. Engineers estimate this AI-assisted approach could cut route-planning time in half.

University of Hawaiʻi Physics-Informed ML (February 19): Researchers unveiled a new algorithm that significantly advances "physics-informed machine learning," allowing AI to adhere to laws of physics while processing complex datasets. Unlike traditional "black box" AI, this approach ensures outputs remain physically plausible even when data is sparse.

UK Fusion Magnet Breakthrough (February): UK engineers at the STEP fusion programme successfully tested pioneering "remountable" high-temperature superconducting magnets with plug-and-socket joints, overcoming a key barrier to commercial fusion reactors—critical for clean AI energy infrastructure.

Darren Aronofsky AI Revolutionary War Series (February 2): Filmmaker's AI venture Primordial Soup released "On This Day... 1776," recreating the American Revolution using Google DeepMind. Each episode drops on the 250th anniversary of the event it depicts, combining AI-generated visuals with SAG-AFTRA voice actors.

Meta News Publisher Deals (February): Meta struck multi-year commercial agreements with news publishers including USA Today, CNN, Fox News, and Le Monde to license content for Meta AI chatbot, enabling real-time news responses with attribution across Facebook, Instagram, WhatsApp, and Messenger.

Google Publisher Opt-Out Complexity (February): Senior Google executive described building a granular opt-out mechanism allowing publishers to exclude content from AI Overviews without affecting traditional search visibility as a major engineering challenge, underscoring tensions between AI features and publisher rights.

UK AI Skills Strategy (February): UK government announced major expansion of national skills strategy, setting a target to upskill 10 million adults in AI skills and competencies by 2030, with new partners including British Chambers of Commerce, NHS, and techUK.

UK Live Facial Recognition Legal Challenge (February): New court case against police use of live facial recognition technology (LFRT) began, challenging Metropolitan Police's deployment of AI-driven biometric identification.

LinkedIn SEO Strategy Overhaul (February): LinkedIn reported that non-brand, awareness-driven B2B traffic declined by up to 60% as AI-powered search experiences reduced clickthrough behavior. The company abandoned traditional SEO metrics in favor of visibility-based measurements centered on mentions, citations, and presence within AI-generated responses.

Curious Refuge AI Film School (February): Online AI filmmaking academy has trained over 10,000 students across 170 countries, becoming a talent pipeline for AI entertainment studio Promise, which acquired it last year. Industry observers describe education as a major opportunity as studios seek AI-literate directors, artists, and storytellers.

BridgeWise-Context Analytics Merger (February): US fintech BridgeWise purchased data processing firm Context Analytics to launch "pAI," a new AI wealth agent that "builds, reviews, and optimizes individual portfolios in a fully personalized, explanatory, and conversational manner."

Basis Series B (February): Founded in 2023, Basis operates an agentic AI platform deploying "long-horizon" agents capable of completing complex account workflows across client services, tax, and audit practices. The Series B funds will accelerate platform development with focus on expanding engineering and machine learning teams.

Commonwealth Credit Union CUSO (February): Partnered with Zest AI to launch CU Lending Collective, a credit union service organization specifically designed to help small credit unions adopt AI-powered lending, offering "an accessible and compliant path" without the typical cost, complexity, or operational burden.

The AI-RAN Alliance reached 132 members worldwide, welcoming Board Members including Qualcomm, SK Telecom, and Vodafone. At MWC 2026, the Alliance presented 33 AI-driven innovation demonstrations and unveiled four industry blueprints highlighting AI integration into Radio Access Networks (RAN) for 5G-Advanced and 6G-ready networks.

February 2026 will be remembered as the month AI companies made the hard choice between idealism and pragmatism—and pragmatism won decisively:

1. Monetization Over Mission: OpenAI's advertising rollout represents the final nail in the "AI for all of humanity" narrative. Despite Altman's framing about democratizing access, the $60 CPM premium and $200,000 minimum investment reveal AI is becoming an advertising platform for deep-pocketed brands, not a tool for the underserved.

2. "Vibe Working" Becomes Real: Anthropic's coining of "vibe working"—the ability for non-technical users to complete complex knowledge work through conversational AI—captures the industry's shift from developer tools to general-purpose assistants. When 16 Claude agents can write a C compiler without human intervention, the implications for enterprise workflows are staggering.

3. The SaaS Apocalypse Fears: The 20%+ drop in the WisdomTree Cloud Computing Fund following Opus 4.6's release signals genuine market concern that AI agents will displace legacy enterprise software. Whether justified or overblown, this fear is driving trillion-dollar valuations for AI companies while legacy SaaS vendors scramble to integrate AI before becoming obsolete.

4. Infrastructure Investment Reaches Bubble Territory: $650 billion in 2026 AI infrastructure spending—a 67% YoY increase—followed by a $950 billion market cap decline suggests investors are questioning whether returns can ever justify the investment. The gap between capital deployed and revenue generated has never been wider.

5. The Reasoning Plateau: Google's doubling of ARC-AGI-2 scores with Gemini 3.1 Pro demonstrates continued progress, but the improvement comes from incremental architectural refinements (.1 releases) rather than paradigm shifts. The industry is optimizing existing approaches rather than discovering new ones.

6. Geopolitical AI Fragmentation: Trump's ban on Anthropic federal use, Chile's Latam-GPT launch, and China/US refusal to sign responsible military AI agreements signal the end of unified global AI governance. Expect regional AI ecosystems with divergent safety standards, training approaches, and deployment restrictions.

7. Model Context Protocol as Standard: With Anthropic donating MCP to the Linux Foundation's Agentic AI Foundation and OpenAI/Microsoft/Google adopting it, 2026 is the year agentic workflows standardize. The "USB-C for AI" finally allows agents to connect to databases, search engines, and APIs seamlessly.

8. Privacy Theater in Advertising: OpenAI's privacy commitments—no advertiser access to chats, no influence on responses—mirror promises from Google and Meta that have eroded over time. The default opt-in for ad personalization and use of past conversations to target ads suggests the privacy protections are more marketing than substance.

Watch for:

February 2026 proved that AI is no longer a technology—it's a business model. The companies that can balance monetization with user trust, navigate geopolitical fragmentation, and deliver measurable enterprise value will define the next decade. The ones that prioritize benchmarks over business outcomes, or idealism over pragmatism, will find themselves disrupted by the very technology they pioneered.

Stay informed, stay critical, and remember: the future isn't built by AI alone, but by the humans who decide how to deploy it. See you in the next release of 'This Month in AI'.