I woke up on Tuesday to find the Salesforce ticker bleeding out like a hit-and-run victim on a Vegas sidewalk. Down 50%. Mid-cap software CEOs are currently screaming into the void of their Pelotons because they’ve finally realized they aren’t selling ‘solutions’—they’re selling digital office furniture in a world that just stopped using offices. This isn’t a correction; it’s a Seldon Crisis appearing 500 years early. The old empire is liquefying.
This isn’t a correction. It’s a category error being repriced in real time.
The Big Picture
AI ate the SaaS moat, and nobody’s sure what comes next.
Jordi Visser called it a “supersonic tsunami.” ARK’s Cathie Wood called it “volatility signals.” Jeremy Lefebvre just sold his biggest position. The narrative is converging: legacy enterprise software is facing an extinction-level event, and the replacement infrastructure is being built by companies that don’t look like software companies at all.
Palantir is up. ServiceNow is hanging on. Salesforce is in freefall. The delta? AI-native architecture. Platforms that were designed to be automated are winning. Platforms designed to sell automation are dying. The market is learning the difference between selling hope in a browser tab and selling compute that actually replaces headcount.
Eighteen sources mentioned Microsoft this week. Not one of them talked about Office. They talked about Azure as the substrate for autonomous agents. They talked about GitHub Copilot as the new IDE. They talked about a future where “enterprise software” means renting GPUs by the millisecond, not seat licenses by the year.
Crypto is having its own reckoning, but it’s quieter.
Ethereum slipped below $2,000. Solana held. The divergence tells the story: networks optimized for agent commerce are holding value. Networks optimized for DeFi speculation are bleeding. Sixteen sources mentioned both ETH and SOL this week, and the through-line was clear — programmable money infrastructure matters when the programs are autonomous.
Chainlink, Polygon, and Fetch.AI all showed up in the convergence data. Not because of token hype. Because enterprises are starting to ask: “If our AI agents need to transact, what rails do they use?” The answer is starting to look like stablecoins on Layer 2s, not Visa.
The physical world is the new alpha.
While everyone panicked about SaaS multiples, the real money quietly rotated into industrial automation, data center REITs, and semiconductor supply chains. SMCI jumped four sources week-over-week. Equinix added five. Cognex — a machine vision company most people have never heard of — went from two sources to eight.
The meta-narrative: AI doesn’t eat services, it eats stuff. While the SaaS clouds evaporate, the ‘Vile Offspring’ are moving into the physical substrate. They are eating the power grids. SMCI and Equinix aren’t just ‘stocks’; they are the raw matter being converted into the Matrioshka brain of the 2030s. If it doesn’t have a cooling pipe attached to it, the market no longer believes it’s real.
Tom Lee’s technician warned of “a choppy, volatile year ahead.” He was talking about equities. He could have been talking about the entire knowledge economy.
The Prime Radiant
The Prime Radiant is flashing red. Five independent vectors have collided on the same coordinate: the ‘Application Layer’ is becoming a vestigial organ. In Foundation terms, we are at the moment where the ‘Mule’ of AI has appeared, and the old charts are being torn to shreds by a probability we didn’t think could exist in this century.
Enterprise AI is no longer a feature — it’s the entire product.
Five sources spanning YouTube and Substack converged on a single idea: enterprises aren’t adding AI to their workflows, they’re replacing their workflows with AI. The investment thesis isn’t “buy the SaaS company that bolts on ChatGPT.” It’s “buy the infrastructure that makes SaaS obsolete.”
All-In Podcast, Cheeky Pint by Stripe, Jeremy Lefebvre, Lex Fridman, and Noahpinion all landed here independently. The phrasing varied — “adoption acceleration,” “automation enablement,” “governance platforms” — but the core insight was identical: the AI layer is becoming the application layer.
What to buy? Not the legacy vendors retrofitting LLMs. The companies selling the rails: cloud compute (MSFT, GOOGL, AMZN), AI-native SaaS (NOW, CRM if they survive), and the governance infrastructure that keeps autonomous agents from going rogue (IBM, weirdly).
Specialized AI compute is the new oil & gas.
Four sources (ARK Invest, Cheeky Pint, Jordi Visser, peterdiamandis) independently flagged the same bottleneck: you can’t run AGI on commodity cloud. You need custom silicon, liquid cooling, and power grids that don’t exist yet.
This is why Nvidia is still up. This is why SMCI is climbing. This is why data center REITs like Equinix are suddenly interesting again. The convergence isn’t on chips, it’s on the entire supply chain required to feed a self-improving model.
The trade: long the specialized compute stack (NVDA, AMD, SMCI, EQIX), short the commoditized cloud resellers who can’t compete on latency or power efficiency.
Legacy SaaS is a turnaround play only if you have AI integration.
ARK and Clouded Judgement both arrived at the same contrarian bet: some legacy SaaS will survive, but only if they gut their UX and replace it with agent-callable APIs. The ones that do this fast enough become acquisition targets. The ones that don’t become case studies.
Salesforce is the test case. If CRM can pivot from “software for sales teams” to “API for sales agents,” it survives. If not, it’s the next Oracle — profitable, irrelevant, and shrinking.
The convergence trade: buy distressed SaaS with strong APIs and cash flow, short the ones still optimizing for human users.
If this is hitting different, subscribe — next week’s convergence drops Monday →
The Board
The meta-story this week: AI is compressing the SaaS layer and expanding the infrastructure layer. The tickers getting multi-source attention fall into three buckets: (1) the hyperscale clouds that host the compute, (2) the crypto rails that let agents transact, and (3) the physical infrastructure that keeps the whole thing from melting. Microsoft and Google are the new operating systems. Ethereum and Solana are the new payment processors. Nvidia and Equinix are the new utilities. Everything else is a bet on which middleware survives.
MSFT · stock · 18 sources · substack, youtube
Outcome-First AI Code Generation Platforms | Enterprise AI Knowledge Acceleration Platforms | Integrated SaaS Moat Plays | Private-Sector Upskilling Platforms
“leaving aside um Microsoft and Google” — Cheeky Pint by Stripe
Sources: a16z, All-In Podcast, ARK Invest, Basis Points, Bowtie Nation, Cheeky Pint by Stripe, Clouded Judgement, Dumb Money Live, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Noahpinion, Not Boring by Packy McCormick, peterdiamandis, RaoulPal, The Compound, The Innermost Loop
ETH · crypto · 18 sources · substack, youtube
Moonshot Superintelligence Application Funds | Middleware APIs for Unofficial Platform Access | AI Governance & Auditability Solutions | Asia-MiddleEast Luxury SUV Supply Chain
“As the price of Ethereum fell below $2,000 each, investors with less than 10,000 ETH.” — Cheeky Pint by Stripe
Sources: a16z, All-In Podcast, ARK Invest, Basis Points, Bowtie Nation, Cheeky Pint by Stripe, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Noahpinion, peterdiamandis, RaoulPal, The Compound, The Innermost Loop, Tim Ferriss, Tom Lee Fundstrat, Visser Labs
NVDA · stock · 17 sources · substack, youtube
AMD/Nvidia Valuation Divergence Play | RSI-Triggered Pullback Entry Strategy for High-Momentum Stocks | AI-Optimized Chipmakers | Technical-Cycle Tactical Rotation
“stopping Nvidia from selling into China.” — Cheeky Pint by Stripe
Sources: All-In Podcast, ARK Invest, Basis Points, Bowtie Nation, Cheeky Pint by Stripe, Dumb Money Live, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Noahpinion, Not Boring by Packy McCormick, peterdiamandis, The Compound, The Innermost Loop, Tim Ferriss, Tom Lee Fundstrat
GOOGL · stock · 16 sources · substack, youtube
Prompt-Engineering SaaS Platforms | Meta AI-Driven Ad Revenue Expansion | Agent-First Platform Infrastructure | Enterprise Prompt-Management & AI-Orchestration Platforms
“leaving aside um Microsoft and Google” — Cheeky Pint by Stripe
Sources: All-In Podcast, ARK Invest, Basis Points, Cheeky Pint by Stripe, Dumb Money Live, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Net Interest, Noahpinion, Not Boring by Packy McCormick, peterdiamandis, The Compound, The Innermost Loop, Tim Ferriss
SOL · crypto · 16 sources · substack, youtube
Programmable Low-Latency Networking Software | Altcoin Relative-Strength Rotation | RSI-Triggered Pullback Entry Strategy | Modular Off-Grid Energy Solutions
“Salana maybe 2,000.” — Cheeky Pint by Stripe
Sources: a16z, All-In Podcast, ARK Invest, Basis Points, Bowtie Nation, Cheeky Pint by Stripe, Jeremy Lefebvre Financial Education, Jordi Visser, Lex Fridman, Not Boring by Packy McCormick, peterdiamandis, RaoulPal, The Compound, The Innermost Loop, Tim Ferriss, Tom Lee Fundstrat
AMZN · stock · 15 sources · substack, youtube
AI-Powered Retail Automation | Infrastructure-as-a-Service for AI Audits | Voice-Driven Prompt Engineering Toolkits | Enterprise Prompt-Management Platforms
“they figured it out and now both Amazon and Walmart are are kings here.” — Cheeky Pint by Stripe
Sources: All-In Podcast, ARK Invest, Basis Points, Bowtie Nation, Cheeky Pint by Stripe, Dumb Money Live, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Net Interest, peterdiamandis, The Compound, The Innermost Loop, Tim Ferriss
SMCI · stock · 10 sources · substack, youtube
Abundance Flywheel Compute Infrastructure | AI Compute & Edge Infrastructure | AI Server Infrastructure Leaders | AI-Infrastructure Expansion Fueled by Monetary Easing
Sources: All-In Podcast, Bowtie Nation, Cheeky Pint by Stripe, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Not Boring by Packy McCormick, peterdiamandis
EQIX · stock · 10 sources · substack, youtube
Abundance Flywheel Compute Infrastructure | AI Compute & Edge Infrastructure | AI Data Center REIT Exposure | AI-Enabled Data Center Infrastructure
Sources: All-In Podcast, Cheeky Pint by Stripe, Jeremy Lefebvre Financial Education, Jordi Visser, Joseph Carlson, Lex Fridman, Noahpinion, peterdiamandis
Momentum Watch
Heating up:
CGNX (Cognex): +6 sources. Machine vision for industrial automation. The robots need eyes.
SWCH (Switch): +5 sources. Data center operator. The quiet infrastructure play everyone suddenly remembers exists.
SOL: +5 sources. Agent-native blockchain. The only Layer 1 still holding narrative momentum.
EQIX: +5 sources. Data center REIT. Physical infrastructure is the new cloud.
Cooling off:
CRM (Salesforce): Holding steady at 11 sources, but the tone shifted. Half the mentions were about survival strategy, not growth.
Legacy SaaS basket: Anything without “AI-native” in the architecture is shedding sources week-over-week.
What We’re Ingesting
Visser Labs — “The SaaS Panic Is Just the Beginning of a Bigger Story”
Jordi’s turbulence model is flashing red. Themes: AI-compressed SaaS valuations, industrial AI moat builders, stablecoin infrastructure. The thesis: software is getting cheaper, hardware is getting more valuable, and money is becoming programmable. Read it if you want to understand why Salesforce is down 50% and Equinix is up.
The Innermost Loop — “Physical Superintelligence”
AI-powered fundamental physics platforms, public benefit corp deep-tech vehicles, shaped-charge physics venture funds. This is what happens when you let an AI recursively optimize material science. The investment angle: companies building the tooling for AI-driven R&D in energy, manufacturing, and materials.
Noahpinion — “How technology has already changed the world in my lifetime”
Decentralized social infrastructure, digital reputation management, enterprise AI automation platforms. Noah’s optimistic, which makes his AI takes worth reading — he’s not selling fear, he’s mapping second-order effects. The through-line: technology compounds faster than institutions adapt, and the gap is tradable.
Net Interest — “The Ackman Complex”
Fee-free capital via insurance platforms, mega-cap time-arbitrage strategies, publicly listed hedge fund shares. Bill Ackman is building a permanent capital vehicle. The meta-lesson: in a world of infinite liquidity and zero rates, the edge is structure, not alpha.
All-In Podcast — “Debt Spiral or NEW Golden Age?”
Asia-MiddleEast luxury SUV supply chain, gold & precious metals inflation hedge, industrial automation & AI robotics. The crew is split: Chamath sees abundance, Sacks sees fiscal collapse, Friedberg sees both. The convergence trade from this episode: long industrial automation, long hard assets, short sovereign debt.
The Closing Note
We’re watching the knowledge economy eat itself in real time. SaaS companies that took twenty years to build are being replaced by API calls that take twenty milliseconds. The people who built those companies are now building the agents that will replace them. It’s recursive, it’s brutal, and it’s probably inevitable.
The irony is that all this abundance — all this AI-driven productivity, all this compute surplus — is making the physical world more valuable. You can’t automate a data center into existence. You can’t prompt-engineer a power grid. You can’t fine-tune a semiconductor fab.
So we’re left with a strange inversion: the digital layer is compressing toward zero, and the industrial layer is expanding toward infinity. The market hasn’t fully priced this yet. It will.
The empire is falling, the agents are transacting on Solana while we sleep, and Hari Seldon has probably predicted the exact second I will realize I’ve lost my left sock in the dryer. The future is a trillion-dollar GPU cluster, and my ankle is still cold. Abundance is a cruel joke.
—
Themis Weekly is written by an anonymous analyst who watches too much YouTube and reads too much Substack. If you found signal in the noise, forward this to someone who needs it. If you want next week’s convergence in your inbox, subscribe below.
Next issue: February 24, 2026
Note [v.2.0]: Consider the modern browser tab—a thin, precarious sliver of memory allocation that currently serves as the final, desperate stronghold of the white-collar workforce. We have spent trillions of dollars building 'Enterprise Resource Planning' tools only to have them end up as a 'favicon' that is ignored 98% of the work day. The AI agent doesn't need the tab; it doesn't even need the monitor. It is a ghost in the machine that makes the GUI look like a cave painting.