🚀 AI + Crypto Convergence: The SpaceX-Anthropic Deal That Chang…

This week marked a turning point in the collision of AI and crypto—a convergence that has been building for over a year but is now erupting across multiple fronts simultaneously. From Solana emerging as the de facto payment rail for AI agents to SpaceX AI leasing massive compute capacity to Anthropic, the pieces are falling into place for what might be the most consequential technology shift since the internet itself.

🧠 Why Wall Street Is Finally Taking AI Agents Seriously

Goldman Sachs released a report declaring that agentic AI could transform AI from a capex-heavy cost burden into a business where usage growth drives margin expansion. The firm's analysis shows that token capacity is projected to increase 24x by 2030—and that's before accounting for enterprise agents, which could push growth even higher.

The takeaway? There will be infinite demand for intelligence. The early innings analogy isn't just rhetoric—current token usage represents a fraction of what's coming as AI agents proliferate across consumer and enterprise applications.

"Token costs are falling, agents are becoming more sophisticated and economically viable, and context windows are getting longer. We are definitely in AGI, the singularity, whatever you want to call it—we are in it right now."

⚡ Solana: The Credit Card for AI Agents

In a significant validation of the thesis discussed here for over 18 months, Anthropic CEO Dario Amodei and Google Cloud VP Tolli confirmed that Solana is becoming "the AI chain." The reason is straightforward: autonomous AI agents need somewhere to transact and pay for services, and Solana's speed and low costs make it the natural infrastructure layer.

Key developments:

Google launched pay.sh, a Solana-based payment system integrated with Google Cloud that connects to powerful APIs for Gemini, BigQuery, and Vertex AI
AI agents use stablecoins on Solana to pay for onchain AI infrastructure, AI-powered apps, and experimentation
Just as stablecoin transaction data showed Solana dominance in crypto payments, AI agent payments are consolidating on the same chain

Meanwhile, Stripe, Amazon, and Coinbase announced a technology partnership to help AI agents pay with stablecoins. Stripe—which powers roughly half of all internet payments globally—is betting that stablecoins will be the payment method of choice for the AI economy.

🔥 The SpaceX AI Master Plan: 5D Chess in the Data Center Wars

Perhaps the most intriguing development of the week was SpaceX AI (the newly merged entity combining XAI and SpaceX) leasing its Colossus 1 supercluster to Anthropic. On the surface, this seems counterintuitive—why would a competitor to Anthropic give them access to world-class compute infrastructure?

The strategic logic becomes clear upon closer examination:

1. Revenue Generation from Idle Assets
Colossus 1, with 220,000 GPUs, was sitting idle as SpaceX AI scaled up Colossus 2 (targeting one million GPUs). Rather than let that capacity go unused, they're monetizing it immediately—likely at premium rates given Anthropic's desperate need for compute.

2. Competitive Positioning Against OpenAI
By empowering Anthropic—which is already outperforming OpenAI in many benchmarks—SpaceX AI accelerates the pressure on OpenAI. Given the ongoing legal battle between Elon Musk and OpenAI, strengthening Anthropic serves as a strategic blow to OpenAI's market position.

3. Intelligence Gathering
By running Anthropic's workloads in their data center, SpaceX AI gains invaluable insights into how a leading AI lab operates at scale. This knowledge becomes critical if SpaceX AI exercises its reported option to acquire Cursor for $60 billion—a move that would give them both the developer tool and the operational playbook to compete directly.

4. Juicing the IPO
SpaceX is reportedly preparing for what could be a $2 trillion IPO. Demonstrating that they can monetize AI infrastructure as a service—in addition to their core space and inference businesses—significantly strengthens the investment thesis.

"The speed at which this conglomerate moves is stunning. XAI has merged into SpaceX to become SpaceX AI—a global telco to power AI, global inference to power AI, and now the world's biggest AI supercluster."

For Anthropic's part, the deal is equally valuable: Claude AI capacity doubled overnight, eliminating the severe rate limits that had been constraining developer adoption.

🎯 The Court Battle Exposing OpenAI's Origins

The ongoing trial in Oakland, California, is revealing explosive details about OpenAI's transformation from nonprofit to for-profit entity. Documents show that Microsoft's $10 billion investment came just weeks after OpenAI restructured—raising serious questions about whether Microsoft had advance knowledge of the corporate structure changes.

The timeline suggests Microsoft conducted extensive due diligence before committing capital, making it highly unlikely they were unaware of the impending shift from 501(c)(3) charity status. Meanwhile, investor Mark Cuban publicly stated that "OpenAI investors will never get their money back"—a damning assessment of the company's financial trajectory despite its market hype.

💎 The Semiconductor Explosion: AMD, Google, and the Trillion-Dollar Capex Cycle

The demand for AI compute reached fever pitch this week, with AMD's stock surging dramatically in after-hours trading following an earnings report that doubled previous forecasts. The company, which traded around $190 five weeks ago, climbed toward $450 as orders for CPUs and custom TPUs vastly exceeded expectations.

AMD CEO Lisa Su's assessment was blunt: demand is insane, and the company cannot manufacture chips fast enough. This supply crunch is forcing hyperscalers to diversify orders across Broadcom, Marvell, and other chipmakers—but even that won't be sufficient.

Google emerged as the clear AI infrastructure leader with a $200 billion commitment for TPUs, to be manufactured by a combination of AMD, Broadcom, and Marvell. The announcement briefly pushed Google's market cap close to $5 trillion, making it the second company (alongside Nvidia) to reach that milestone.

The bears immediately pointed to concerns: Google's capex is consuming free cash flow, and the stock is trading at 133 times free cash flow—a valuation not seen since near-IPO levels. But this criticism misses the point entirely. When a single customer order ($200 billion from Anthropic and SpaceX AI) approaches the total capex budget, the returns become obvious.

"If they spend $250 billion and have one $200 billion order from one client, that's pretty good. The demand for this stuff is infinite and they know it."

Other chip plays saw similar explosions: Marvell and Astera Labs both doubled in the past month—moves not seen since the dot-com era, but this time backed by genuine demand rather than speculative fervor.

🏗️ The Trillion-Watt Future: Musk's Terafab and the Battle for Inference

SpaceX AI and Tesla are jointly pushing for trillion-watt compute power through proprietary "terafabs" producing 2-nanometer AI chips. The strategic focus? Inference over training.

As the analysis emphasizes, inference is everything. Models only need to be trained once, but they must perform inference billions of times per second—every time an autonomous vehicle encounters an edge case, every time a humanoid robot navigates an environment, every time an AI agent executes a transaction.

This insight drives the thesis that whoever controls inference compute controls the AI economy. If Tesla and SpaceX AI can manufacture superior inference chips at scale, they could disrupt even Nvidia and AMD—or more likely, all players will coexist because demand will be that enormous.

🤖 The Robotics Inflection Point: Humanoids Are Here

The most significant leading indicator for the robotics revolution comes from Hugging Face, the world's leading platform for AI datasets. In 2022, the top category was text classification. By 2023-2024, it was text generation.

But in 2025? Robotics datasets have exploded to become the number one category by far—despite not existing at all in 2023. This is not a future trend; robotics is already here.

The reason Tesla is tight-lipped about Optimus 3 is strategic: they don't want competitors (particularly Chinese manufacturers) to reverse-engineer the design. But make no mistake—humanoid robots represent the largest product category that will ever hit the planet, dwarfing even chips and autonomous vehicles.

"2026 is going to be a transformative year. 2027 will be off the hook. Those fearing 2026—yeah, we had a nice dip in early 2026, great buying opportunities, but from now on it should be pretty good."

Tesla's robotaxi deployment in Austin, Texas, continues its vertical trajectory, with unsupervised vehicles now representing roughly 70% of the fleet. At the current pace of expansion, 1,000 unsupervised vehicles by year-end is easily achievable—potentially reaching 2,000 or even 5,000. Once that threshold is crossed, Wall Street confidence will solidify, and capital will flood in.

🏥 The Surprise AI Winner: Tether's Medical AI

In one of the week's more unexpected developments, Tether—the stablecoin issuer—revealed a medical AI model that is outperforming Google's MedGemma while running on an iPhone. The model is 16 times smaller than Google's yet delivers superior results.

This stunner demonstrates how much disruption potential exists when unexpected players enter the AI space. That said, Grok remains the number one model for medical and health questions—a fact worth remembering for anyone seeking a second opinion on a diagnosis or medical concern.

⚠️ The Orwellian Corner: Labor Disruption and Social Unrest

Coinbase CEO Brian Armstrong announced 14% staff layoffs, citing efficiency gains from AI. The reasoning was stark: engineers can now ship in days what used to take months, and even non-technical staff can ship production code with AI.

This is the beginning of a broader labor transformation that venture capital firm A16Z analyzed in a recent report. AI productivity is creating labor surpluses, which in turn is spawning new job categories: Pilates instructors, athletes, coaches, dog walkers, exam prep tutors, nail salons, and other in-person service roles.

But not everyone will transition smoothly. In South Korea—a country known for its reserved, polite culture—citizens are already taking to the streets demanding their share of AI profits. If this is happening in South Korea, the potential for civil unrest in other regions is significant.

"Can you imagine what people will do in places like Portland, Seattle, and parts of New York when they discover that AI is ripping through the world as they know it? The next five years will be very bumpy."

📊 Investment Thesis: Positioning for the AI Stack

The strategic positioning breaks down into three tiers:

Defensive Strength:

Marvell
Astera Labs
AMD
Google (via massive TPU orders and cloud infrastructure)

Core Tested:

Nvidia (still the gold standard despite competition, with Elon Musk confirming SpaceX AI will buy as many Nvidia chips as they can produce)

High Conviction:

SpaceX AI (terafab and inference dominance)
Tesla (robotaxis, humanoids, and custom inference chips)

The overarching thesis: This is a once-in-a-lifetime revolution running through 2030 and beyond. Token usage will grow 24x. Robotics is exploding. Humanoids are no longer the future—they're the present. And the firms building full-stack AI infrastructure are positioning for trillion-dollar outcomes.

🎬 The Bottom Line

The worlds of AI and crypto are no longer separate. Solana is the credit card for AI agents. Stablecoins are the payment rail. SpaceX AI is running 5D chess in the data center wars. The companies that control inference compute, semiconductor supply, and AI agent infrastructure will dominate the next decade.

This isn't speculation. The orders are real. The deployments are happening. The talent wars have begun. Forewarned is forearmed. Build defensive positions. Get exposure across the AI stack. And buckle in—because the revolution is already here.