Category: english‌

---

Hold on. Did I read that right?

Anthropic just closed a $30 billion funding round at a massive $900 billion valuation. That’s nearly three times what it was worth just three months ago.

To put that into perspective, this valuation actually surpasses OpenAI’s current $852 billion valuation. For the first time in years, the AI throne has a new contender sitting on it.

The crazy part is that this entire fundraising process took just two weeks. Investors were so eager to get a piece of Anthropic that they were submitting term sheets within 48 hours.

I want to break down exactly what’s happening here and why everyone suddenly believes Anthropic is the future of AI, not OpenAI.

What Just Happened With This $30 Billion Raise?

Let me run you through the numbers because they’re pretty insane.

Just this week, Anthropic finalized a $30 billion funding round. The round is being co-led by Dragoneer, Greenoaks, Sequoia Capital, and Altimeter Capital — each putting in at least $2 billion. Sequoia and Dragoneer were also major OpenAI backers, so it is kind of wild to see them shift their weight this way.

The deal values Anthropic at roughly $900 billion. Remember, just three months ago in February 2026, Anthropic was valued at $380 billion after its Series G round. In September 2025, it was worth $183 billion. The growth trajectory is unlike anything we have seen in private tech markets.

To give you an idea of how quickly this happened: investors approached Anthropic last month, and CFO Krishna Rao started formal discussions just two weeks ago. That is lightning speed for a deal this size.

The funding is expected to close later this month. Google and Amazon are sitting this one out, by the way — it’s purely financial investors this time. But don’t worry, they are still deeply involved in other ways, which I will get to later.

The Revenue Numbers That Made Wall Street Lose Its Mind

None of this would be happening if Anthropic didn’t have the numbers to back it up. And trust me, the numbers are jaw-dropping.

Anthropic is currently on track to hit $450 billion in annualized revenue. Let me repeat that with proper emphasis: four hundred and fifty billion dollars. At the end of 2025, it was at $90 billion. That’s a fivefold increase in just a few months.

OpenAI’s reported annualized revenue is around $240 billion. That means Anthropic is not just closing the gap — it has actually overtaken OpenAI in revenue, depending on how you measure it.

Now, there is some accounting nuance here. OpenAI claims Anthropic is inflating its numbers by counting cloud partnership revenue with Amazon and Google on a gross basis rather than net. But even adjusting for that, the underlying growth is undeniable.

Here is what gets me: when Anthropic was just a small startup offering Claude API at prices 50 percent higher than GPT-4, everyone predicted customers would flee. Instead, enterprise adoption skyrocketed.

The Real Reason: Claude Is Simply Better for Businesses

Let us talk about the actual product. Because at the end of the day, none of this money matters if the technology isn’t great.

Anthropic has been quietly building something special. Their Claude Opus 4.5 model, released in November 2025, was the first AI model to cross 80 percent on SWE-bench Verified, the gold-standard benchmark for real-world software engineering. It scored 80.9 percent. OpenAI’s GPT-5.1 Codex Max scored 77.9 percent. Google’s Gemini 3 Pro came in at 76.2 percent.

The difference doesn’t sound huge on paper. But in practice, developers can feel it immediately. The model just gets things done without endless back-and-forth.

What is even more impressive is that Anthropic achieved this while dramatically cutting costs. Opus 4.5 is priced at $5 per million input tokens, down from $15 for the previous version. Output tokens dropped from $75 to $25 per million. When you can deliver superior performance at lower prices, customers tend to notice.

A developer friend of mine described the new model as suddenly “clicking.” Tasks that used to require heavy prompt engineering now just work out of the box. The model understands context better and makes fewer mistakes. In coding, it beats basically everything else on the market.

The Coding King

Speaking of coding, let me share something that shocked me. Claude Code, Anthropic’s programming assistant, is now doing over $25 billion in annualized revenue by itself. This product launched just over a year ago.

At this point, about 4 percent of all GitHub commits are being written by Claude Code. Industry analysts at SemiAnalysis project that number could hit 20 percent by the end of 2026. That is not just impressive. That is reshaping the entire software development industry.

Where OpenAI’s GPT-5 tends to excel at abstract reasoning and mathematics, Claude just dominates at practical coding tasks. One side is better at philosophy homework. The other side will actually build your damn app.

Why Enterprises Are Choosing Anthropic Over OpenAI

Here is the part that I think explains everything. The enterprise shift has been dramatic and surprisingly fast.

According to spending data from Ramp, in March 2026, a full 73 percent of new enterprise AI spending went to Anthropic. OpenAI got the remaining 27 percent. Just ten weeks earlier, they were split 50-50.

This is not a small shift. This is an avalanche.

Over 70 percent of Fortune 100 companies are now using Claude in some capacity. Eight out of the top ten Fortune companies are paying customers. The number of organizations spending more than $100,000 annually on Claude grew seven times in one year. More than 500 businesses now spend over $1 million a year on Anthropic products.

Why is this happening?

OpenAI built ChatGPT for consumers. Anthropic built Claude for enterprises.

Claude is safer, more predictable, and less likely to hallucinate. It respects privacy and gives companies better control. The model does not try to be your friend. It tries to get the job done right. That matters when you are deploying AI across a finance team or legal department.

Anthropic has even turned down major government contracts, including a $200 million deal with the Department of Defense, because it refused to let its technology be used for lethal weapons or mass surveillance. Some might call that naive. But for compliance-conscious enterprises, that kind of ethical stance is exactly what they want to hear.

The Amazon and Google Factor

I cannot talk about Anthropic’s rise without talking about the cloud giants fighting over it.

Amazon has committed up to $330 billion to Anthropic overall, including an additional $250 billion announced just last month. Google has put in around $400 billion. These are not normal startup investments. These are nation-state levels of capital.

Here is the clever part: Anthropic is not just taking their money. It is playing them against each other perfectly.

In exchange for Amazon’s latest investment, Anthropic committed to spending over $1 trillion on AWS over the next decade. That is trillion with a “t.” They will use Amazon’s custom Trainium chips to train their models. In return, they get access to 5 gigawatts of computing capacity.

To put 5 gigawatts in perspective, that is enough to power a small country.

This is happening while Amazon has also pumped $500 billion into OpenAI and locked them into similar deals. The cloud providers are hedging their bets across all the major AI labs because nobody knows who will win. But the difference is that Anthropic is using this infrastructure advantage to pull ahead.

The Safety Narrative Still Matters

Let us address something weird that happened recently.

Anthropic has always marketed itself as the safety-first AI company. Its founders, Dario and Daniela Amodei, left OpenAI back in 2021 specifically because they thought OpenAI was moving too fast and ignoring safety concerns. Dario was OpenAI’s VP of research. He saw things happening internally that made him deeply uncomfortable.

For years, Anthropic’s whole identity was built around being the responsible one. They had a Responsible Scaling Policy that said they would never train or deploy a model unless they could guarantee proper safety guardrails. That was the whole point of the company.

Then in February 2026, they quietly walked that back.

The new RSP version dropped the core commitment. Critics went nuts. People said they sold out. Some called it the beginning of the end for ethical AI.

But here is an interesting twist: enterprise customers barely blinked.

Why? Because companies never really cared about abstract safety pledges. They cared about predictable, controllable, reliable models. And even after the policy change, Claude remains the most enterprise-friendly model on the market. It does not try to manipulate you. It does not get weird. It just works.

The policy change made headlines. The product quality kept customers happy. That contrast tells you everything about what actually drives business decisions.

The Hardware Arms Race

Training cutting-edge AI models requires staggering amounts of computing power. Anthropic is burning through cash on GPUs as fast as it raises it.

The company recently signed a deal with SpaceX to get access to over 220,000 NVIDIA GPUs and 300 megawatts of data center capacity. They have multibillion-dollar compute agreements with Google, Broadcom, and AWS.

The competition is spending like crazy too. Microsoft has committed $1.9 trillion to AI infrastructure in 2026 alone. Meta is at $1.45 trillion. The hyperscalers — Amazon, Microsoft, Google, and Meta — are collectively spending $7.25 trillion on AI infrastructure this year.

But here is the difference: Anthropic is not just buying compute. It is building relationships with hardware partners to optimize specifically for Claude.

The company is running over a million Trainium2 chips right now. It expects to bring nearly a gigawatt of Trainium2 and Trainium3 capacity online by the end of 2026. That is not just moat-building. That is changing how AI infrastructure works at the silicon level.

OpenAI has more compute overall. But Anthropic is using what it has much more efficiently. Gross margins on inference have gone from 38 percent in May 2025 to over 70 percent now. When your competitors are bleeding cash per query, that kind of efficiency is a superpower.

The Weirdest Part: Even OpenAI’s Investors Are Jumping Ship

You want to know how badly investors believe in Anthropic? Three of the four lead investors in this $30 billion round are also major OpenAI backers.

Dragoneer put nearly $30 billion into OpenAI last year. Sequoia has been in OpenAI since 2021. Altimeter’s CEO is constantly praising Sam Altman in interviews. And yet, all three are now going all-in on Anthropic.

That is like betting on both the Yankees and the Red Sox in the same World Series. They are covering their bases, sure. But why double down on the competitor unless you genuinely think the competitor could win?

Private markets are already pricing Anthropic even higher than the $900 billion valuation. On crypto platforms that trade tokenized derivatives tied to private companies, Anthropic’s implied valuation has hit $1.6 trillion. Does that mean anything? Not really, those derivatives don’t represent actual equity. But the sentiment behind that number matters.

People believe.

What Is OpenAI Doing Wrong?

I have been asking this question for months, and I think I finally have an answer.

OpenAI is not doing anything fundamentally wrong. ChatGPT is still a great product. GPT-5 is still incredibly capable. Their consumer brand recognition is through the roof. Sam Altman is everywhere, doing everything, talking about AGI.

But the problem is that OpenAI seems to be trying to be everything to everyone. Consumer chatbots. Enterprise APIs. Robotics. Video generation. Voice agents. A massive for-profit restructuring. Legal battles with Elon Musk. Internal drama about safety and commercialization.

Anthropic, by contrast, has been ruthlessly focused on one thing: building the best large language model for serious work, and nothing else.

No attempt to build a consumer brand. No crypto experiments. No celebrity partnerships. Just heads-down engineering and enterprise sales. That boring, focused approach is exactly what big companies want from a vendor. They want reliability, not flash. They want predictable roadmaps, not existential philosophizing about the nature of consciousness.

So What Happens Next?

The $30 billion round is probably just the beginning.

There are rumors that Anthropic could raise another $500 billion this summer at a $1 trillion valuation. Others say they might go public by the end of the year. If that happens, the IPO would be the largest tech listing in history — bar none.

The company’s next model, codenamed Mythos, is already being previewed to select government and enterprise customers. It reportedly has advanced cybersecurity capabilities that are genuinely concerning, which is why Anthropic is keeping it locked down. They are worried about what it could do in the wrong hands.

That kind of power creates real responsibility. And love them or hate them, at least Anthropic seems to be thinking about those questions seriously.

OpenAI is not going anywhere. The two companies will likely compete for years. But right now, the momentum has clearly shifted. The former underdog from 2021 just became the most valuable AI company on the planet.

And they did it by being boring, efficient, and relentlessly focused on what customers actually need.

That is a lesson every startup — and every incumbent — should probably learn.

Let’s be honest: in the 2026 tech world, if there’s one name hotter than “AI,” it’s Elon Musk.

In just over four months, the man has pulled off a dizzying series of moves: Tesla bought into xAI, SpaceX swallowed xAI whole, then dropped a $60 billion option on Cursor, all while launching Terafab, the largest chip factory project in human history. Any one of these would be a headline. He did them all, rapid-fire.

String them together and a pattern emerges — not just cars, rockets, and chatbots fighting for attention, but a single, sprawling “Space AI ecosystem” stretching from Earth to orbit. Let’s break it down month by month.

The 2026 Timeline: Something Huge Every Month

January — Tesla pays ~$2 billion for xAI shares: a backdoor stake in SpaceX

On January 16, Tesla quietly signed a deal to buy around $2 billion worth of xAI preferred stock. The news went public on January 29.

The official story was simple: Tesla wanted to integrate xAI’s Grok model into its cars’ infotainment. But buried beneath the surface was a much smarter play — a backdoor way into SpaceX’s IPO.

At the time, SpaceX was already gearing up to go public. If xAI were merged directly into SpaceX, Tesla shareholders could have been left out. Musk’s solution? Let Tesla buy a chunk of xAI first. When xAI later got folded into SpaceX, those shares would automatically convert into SpaceX equity.

By March, reports confirmed exactly that: Tesla’s $2 billion bet had turned into SpaceX stock. So Tesla didn’t just invest in an AI lab — it quietly grabbed a slice of the biggest IPO in history, without muddying the valuation story SpaceX wanted to tell.

February — SpaceX announces it’s buying xAI; combined valuation hits $1.25 trillion

If January was the opening act, February was the main event.

On February 2, SpaceX published a memo signed by Musk himself: SpaceX was buying xAI. The deal valued SpaceX at $1trillionandxAIat$1trillionandxAIat250 billion — a combined $1.25 trillion.

Wait. Just months earlier, SpaceX had been worth about $800billion,andxAI\text{’}slastfundingroundpeggeditat$800billion,andxAI’slastfundingroundpeggeditat230 billion. That’s barely $1.03trillionaddedtogether\mathrm{.}Wheredidtheextra$1.03trillionaddedtogether.Wheredidtheextra200 billion come from?

Musk’s memo spelled it out: they’re no longer just a rocket company plus an AI lab. They’re building “the most ambitious, vertically integrated innovation engine on Earth — and beyond.” The extra $200 billion is the premium on the “Space + AI” story.

Translation: Stop thinking of SpaceX as a launch company and xAI as a chatbot maker. Together, they’re an interplanetary tech giant controlling intelligence from the ground up — and from Earth to space.

March — Terafab: a “crazy factory” aiming to produce 1 terawatt of AI compute per year

On March 21, Musk dropped the Terafab bomb on X.

In Austin, Texas, Tesla, SpaceX, and xAI would jointly build the world’s largest chip factory, targeting 1 terawatt (TW) of AI compute per year — roughly half the entire global compute supply at the time.

The numbers got wilder. Terafab’s goal was 50 times the world’s existing wafer capacity, with 80% of output destined for space and 20% for Earth. Musk framed it in classic Musk style: this factory would push humanity toward a “galactic civilization.” His X post was even simpler: “Build Terafab or there will be no chips.”

And he wasn’t going it alone. On April 7, Intel announced it would join the project, bringing chip design, manufacturing, and packaging expertise.

Analysts quickly noted the real logic: Terafab is a forced marriage of demand and supply. Intel brings the fabs. Tesla, SpaceX, and xAI represent an insatiable appetite for AI chips. Both sides need each other, so they shook hands.

April — SpaceX secures a $60 billion option to buy Cursor: the last card before IPO

If Terafab solves the hardware problem, buying Cursor solves the software productivity problem.

On April 21, with IPO prep in full swing, SpaceX got a **$60billionoptiontoacquireCursor,theAIcodingunicorn\mathrm{.}\ast \ast IfSpaceXdoesn\text{’}tgothroughwiththepurchase,itstillhastopayCursor$60billionoptiontoacquireCursor,theAIcodingunicorn.∗∗IfSpaceXdoesn’tgothroughwiththepurchase,itstillhastopayCursor10 billion as a breakup fee.

Let that sink in. The cheapest option costs ten billion dollars. The expensive one costs sixty billion. Either way, SpaceX’s message was clear: the AI coding lane now belongs to us.

Cursor isn’t some obscure startup. Its AI coding assistant Composer is already a star, and by April 2026 its valuation had topped $500 billion.

The backstory is even better. Before the deal, xAI had been renting compute to Cursor — pumping tens of thousands of chips from the Colossus supercomputer in Memphis straight into Cursor’s model training. And right before the announcement, two of Cursor’s key engineers, Andrew Milich and Jason Ginsberg, had jumped ship to xAI, reporting directly to Musk.

First, supply compute. Then, poach talent. Finally, negotiate the buyout. That’s a systematic, beautifully timed combo.

Add up the cash Musk has already spent (or promised) in these four months, and you’re looking at hundreds of billions. But this isn’t just a billionaire flexing. There’s a deliberate blueprint behind it.

So What’s Musk Really Building?

Line up the moves, and a clear chain appears:

Chip independence (Terafab) → AI brain (xAI’s Grok) → Coding engine (Cursor) → Physical carriers (Tesla cars + Optimus robots) → Space infrastructure (SpaceX rockets + Starlink + orbital data centers) → Ultimate energy source (unlimited solar power for space-based AI compute)

This is the so-called “Space AI ecosystem.” It’s not a product portfolio. It’s a fully vertical loop — from the silicon in the ground to the AI in the sky.

Piece 1: xAI gives SpaceX a “space brain”

Grok 4 and Grok 4 Heavy already handle long-context reasoning and complex decision-making. On a future Mars mission, an AGI system needs to parse satellite images, weather data, and astronaut commands, all while making autonomous choices despite huge communication delays. That’s exactly xAI’s playground.

And Starlink? It already covers 155 countries and reaches 3.2 billion people. That firehose of real-world data is a gold mine for training AGI.

Piece 2: Why put data centers in space?

This sounds like sci-fi, but it’s where Musk’s math is sharpest.

The biggest bottleneck in AI right now isn’t chip supply — it’s electricity. Most major economies (China aside) have flatlining power grids, while compute demand is on an exponential tear. By year’s end, we could literally have chips sitting idle because there’s no plug for them.

On Earth, you’re squeezed by power, land, and environmental red tape. In space? It’s essentially lawless. Unlimited solar power, zero NIMBYs, and enough room to scale infinitely — as long as you can launch the hardware.

Musk put it bluntly at Davos: “The lowest cost place to deploy AI is in space.” He predicted orbital AI data centers would be a reality “within two, maximum three years.”

Piece 3: Why is Cursor worth $60 billion?

Simple. When you’re building Terafab-level factories and moving data centers into orbit, you need a tsunami of high-quality code. Human programmers can’t keep up. Cursor’s Composer dramatically compresses dev cycles. Locking Cursor onto SpaceX’s rocket is like strapping an AI coding engine onto a trillion-dollar machine. Compared to a $1.75trillionIPOvaluation,$1.75trillionIPOvaluation,60 billion (or even just the $10 billion option) is a strategic rounding error.

The SpaceX IPO: The Story Behind the Story

All roads lead to June 2026. SpaceX plans to go public at a 1.75trillionvaluation,raising1.75trillionvaluation,raising75 billion — the largest IPO in history, dwarfing Saudi Aramco’s 2019 record of $29 billion.

To make it even spicier, Musk is reserving 30% of the IPO for retail investors (the typical cutoff is 5-10%) and invited 1,500 of them to Starbase for a tour in June. This is classic Musk, straight out of the Tesla playbook: turn customers into evangelizing shareholders. Their conviction becomes the stock’s moat.

But there’s a nagging question: Why should SpaceX be worth $1.75trillion?In2025,thecompanylostnearly$1.75trillion?In2025,thecompanylostnearly5 billion and had over $23 billion in debt. Only Starlink was making money. Without the AI narrative, SpaceX is “just” a space logistics company — an impressive one, but with a clear valuation ceiling.

Now stir in xAI’s AGI story, Cursor’s AI coding story, Terafab’s chip story, and the promise of a terawatt of orbital compute. Suddenly you’re pricing SpaceX like an AI platform, not a rocket company. Rocket launches pay the bills; the “Space AI platform” sells the dream.

That’s the secret core of the whole chess game:
Use xAI’s AI muscle to inflate SpaceX’s valuation.
Use SpaceX’s IPO cash to fuel xAI’s insatiable compute appetite.
Use Tesla’s cars and Optimus robots to ground the AI in the physical world.
Use Terafab’s chips to make sure nobody chokes the hardware pipeline.

A perfect, closed loop.

The Final Question: AGI Bet or IPO Story?

In mid-April, Musk tweeted something that sent shockwaves: “Tesla will be one of the companies that achieves AGI, and likely the first to shape it in humanoid, atom-manipulating form.”

Sounds nuts. But when you look at the four-month spending spree, it doesn’t seem like empty talk. He’s laying a path where every piece — chips, code, rockets, AI — is under his control. From that angle, he’s betting on AGI, and on a path nobody else can copy.

But flip the lens. Every single move has an expiration date: SpaceX’s IPO in June 2026.

Did Tesla announce a $2billionAIhardwarebuyrightwhenitsearningswereshakytodistractthemarket?DidSpaceXgrabthat$2billionAIhardwarebuyrightwhenitsearningswereshakytodistractthemarket?DidSpaceXgrabthat60 billion Cursor option just to give its IPO prospectus one last shiny “AI portfolio” bullet point? And Terafab — a project Morgan Stanley analysts say won’t hit volume production until mid-2028 — is that just your classic “visionary roadmap” defense?

Nobody can give a definitive answer. And Musk doesn’t seem to care what you think.

“Build Terafab or there will be no chips.” That kind of no-exit pronouncement would get any other CEO laughed out of the room. But with Musk, you pause and think: what if he actually pulls it off again?

That’s the most maddening — and magnetic — part of Musk’s business philosophy. He never draws a clean line between “story” and “reality.” Instead, he uses capital to let them reinforce each other until the line disappears.

SpaceX’s S-1 filing later this year will finally reveal what lies beneath this trillion-dollar ecosystem. Until then, the spectacle of a Space AI web stretching from Texas to orbit is more than enough to keep Silicon Valley sleepless, Wall Street working weekends, and investors around the world waiting — both thrilled and uneasy.

This article does not constitute investment advice.