What’s Coming, What It Means, and What You Need to Do – William Stickevers

By Peter H. Diamandis, MD
Founder & Exec Chairman, XPRIZE

When I interviewed Elon Musk earlier this year, he described what’s ahead as a supersonic tsunami of converging exponentials.

That phrase stuck with me because it captures the moment perfectly.

What we are witnessing is not linear progress. It is not business as usual. It is not another incremental upgrade cycle.

We are watching three exponential curves hit inflection points at the same time: compute scaling, model capability, and infrastructure deployment. And when exponentials converge, they do not produce marginal change. They produce phase transitions.

Over the last several weeks, the evidence has become impossible to ignore.

The tsunami is not coming.

It is already here.

Let’s look at what is happening, what it means, and what you need to do now.

The Numbers Proving the Tsunami Is Real

Before getting into the technological breakthroughs, start with the raw economic signal.

What is happening in AI revenue right now is unprecedented in business history.

Anthropic reached $14 billion in annualized revenue in February 2026, up from $1 billion just 14 months earlier. That figure has since climbed past $19 billion, more than doubling from $9 billion at the end of 2025. There is no real precedent for this in B2B software. Not Slack. Not Zoom. Not Snowflake. Nothing.

To put that in perspective, Anthropic’s monthly revenue run rate is now roughly $1.6 billion per month — more than Snowflake generates in an entire quarter — and it is still accelerating. Anthropic is reportedly projecting as much as $70 billion in revenue by 2028.

OpenAI reached $25 billion in annualized revenue by the end of February 2026, up from $21.4 billion at year-end 2025, with full-year 2025 revenue at $13.1 billion.

Both companies are now valued in the hundreds of billions. Anthropic sits around $380 billion following its $30 billion Series G. OpenAI’s most recent private round, in February 2026, reportedly valued it at approximately $730 billion, with an IPO potentially targeting the $1 trillion mark.

Then there is Jensen Huang.

He recently finalized a $30 billion investment in OpenAI and a $10 billion investment in Anthropic, while signaling to investors that these will likely be Nvidia’s last private investments in either company because both are heading toward public markets.

Think about that.

The CEO of Nvidia — the person with perhaps the clearest line of sight into real AI infrastructure demand on Earth — just placed $40 billion in final pre-IPO bets on the two companies at the center of the wave.

That is not a gesture. That is a signal.

What Is Actually Driving the Revenue

This is where it gets more interesting.

This growth is no longer being driven primarily by IT budgets. It is being driven by a much larger pool of spending: labor budgets.

Models like Claude and GPT-5 have crossed a threshold. They are no longer just software tools. They are beginning to compete directly with human work.

Companies are not buying AI simply to replace servers or automate workflows at the margins. They are buying it to augment, compress, and eventually displace labor costs.

The clearest breakthrough use case is coding.

Anthropic’s agentic coding product, Claude Code, now reportedly exceeds $2.5 billion in run-rate revenue, having more than doubled since the start of 2026. Business subscriptions have quadrupled. Enterprise adoption now accounts for more than half of Claude Code’s total revenue.

That matters because software engineering has always been a core bottleneck in company building.

Startups can never hire enough engineers. The Fortune 500 struggles to attract top engineering talent because so much of it flows into Silicon Valley. But now intelligence itself is becoming purchasable on demand.

Metered. Elastic. API-accessible.

No recruiting. No vetting. No retention risk. No equity grants.

Just intelligence as a utility.

Consumers pay $20 a month. Enterprise power users pay $200 a month. Large companies are spending millions per year because the ROI is already there.

This is the real transition: intelligence is moving from a scarce human input to an abundant digital service.

The Infrastructure Equation

Now look beneath the applications and revenue growth to the infrastructure layer.

This is the part that still is not being discussed loudly enough.

The five largest U.S. hyperscalers — Microsoft, Alphabet, Amazon, Meta, and Oracle — have collectively committed to roughly $690 billion in capital expenditure in 2026 alone, nearly double 2025 levels. Most of that is aimed at AI compute, data centers, and networking.

Global AI spending is forecast to reach $2.5 trillion in 2026, a 44% increase over 2025, according to Gartner.

Data centers. GPUs. Power generation. Networking. Chip fabrication.

This is not just another tech cycle.

It is the largest infrastructure buildout in the history of technology, and it is happening by a wide margin.

A useful rule of thumb in this world is this: roughly $50 billion per gigawatt of infrastructure, and approximately $10 billion of annual revenue per gigawatt.

Energy equals intelligence.

On a recent earnings call, Jensen Huang estimated that between $3 trillion and $4 trillion will be spent on AI infrastructure by the end of the decade.

This is not hype.

This is capital deployment on a scale that rewrites what is possible.

When you are spending $50 billion on a single data center and generating $10 billion a year from it, you are not just building a product. You are building a new economic substrate.

You are building the electricity grid of the 21st century.

The tsunami is here.

The only question is whether you are building on top of the wave — or getting buried by it.

AI: The Capability Jump

Those revenue numbers are not being driven by narrative alone. They are being driven by real breakthroughs in capability.

Start with this: neuromorphic chips have reportedly solved complex physics simulations at 1,000x better energy efficiency than conventional supercomputers.

That is not a 10% gain. That is a three-orders-of-magnitude jump.

And when computing becomes that cheap, the impact is not simply that existing tasks get faster. Entirely new classes of application become economically viable.

Drug discovery can move from weeks on supercomputer clusters to hours on desktop-scale hardware. Climate modeling that once required national labs can begin to run on university systems. Real-time protein folding for personalized cancer treatment starts to look practical.

This is classic exponential dynamics: dematerialization, demonetization, and democratization accelerating into disruption.

At the same time, China’s DeepSeek is pushing next-generation models through Huawei and Cambricon rather than relying on U.S. chips. The AI race is now unmistakably multipolar.

NVIDIA is releasing new systems aimed at bringing reasoning into autonomous vehicles and robotics — what some are already calling a “ChatGPT moment for the physical world.”

That is the broader shift:

AI is moving from virtual to physical, from U.S.-dominant to globally distributed, and from expensive to radically cheaper — all at once.

And the revenue confirms this is no longer experimental.

This is already being deployed in production by major corporations, defense organizations, and frontier technology firms for mission-critical use cases.

Energy: The Bottleneck Is Real — and So Is the Response

Of course, there is an obvious constraint: power.

Those $50 billion data centers require enormous amounts of electricity, and the grid was never designed for this level of load growth.

Global electricity demand from data centers is expected to more than double by 2030, reaching roughly 945 terawatt-hours — about the equivalent of Japan’s annual electricity consumption.

In the United States, data centers are expected to account for nearly half of all electricity demand growth through 2030. Electricity demand from AI-optimized data centers alone is projected to more than quadruple.

Lawrence Berkeley National Laboratory estimates that U.S. data center electricity demand could rise from 176 TWh in 2023 to between 325 and 580 TWh by 2028, representing as much as 12% of total U.S. electricity consumption.

The grid was not built for this.

Interconnection queues are backed up for years. Transmission permitting can take a decade. And the generation capacity required does not yet fully exist.

In northern Virginia alone, a 2024 voltage fluctuation triggered the simultaneous disconnection of 60 data centers — a warning shot of what grid strain at scale can look like.

But here is the key point:

The bottleneck is being attacked from every direction at once.

Fusion Is Converging Faster Than Expected

Fusion is no longer a permanent-future story. It is becoming a compressed-timeline story.

China’s EAST reactor recently surpassed a major plasma-density barrier long considered out of reach. France’s WEST tokamak sustained plasma for more than twenty minutes. EAST maintained high-confinement plasma for nearly eighteen minutes — exactly the kind of stability required for eventual commercial operation.

Meanwhile, private-sector fusion is moving faster than most people realize.

Commonwealth Fusion Systems has raised nearly $3 billion, including backing from Nvidia and Google, with the long-term goal of a 400-megawatt power plant — enough to power roughly 280,000 average U.S. homes. Its SPARC demonstration machine is expected to produce first plasma in 2026 and pursue net fusion energy shortly thereafter, setting the stage for ARC, a grid-connected power plant targeted for the early 2030s.

Helion Energy has also begun construction on its first commercial fusion plant, designed to deliver power directly to Microsoft data centers starting in 2028.

Private fusion investment has grown to $10.6 billion between 2021 and 2025, while the number of private fusion companies has more than doubled from 23 to 53.

The old joke was that fusion was always 30 years away.

That joke is aging badly.

“Fusion in 30 years” is increasingly becoming fusion this decade.

And in a final layer of irony, AI itself may be helping accelerate the plasma physics research needed to make fusion practical.

The same technology creating the energy crisis may also help solve it.

The Wild Card: Tesla’s Terafab

Then there is Tesla.

On March 14, 2026, Elon Musk posted on X that the “Terafab Project launches in 7 days” — pointing to March 21.

What is Terafab?

Musk first outlined the concept at Tesla’s 2025 shareholder meeting, describing a chip fabrication facility on the scale of TSMC’s largest plants. During Tesla’s January 2026 earnings call, he made the rationale explicit: Tesla would need to build a domestic TeraFab spanning logic, memory, and packaging to avoid hitting a hard ceiling on chip supply within three to four years.

The vision is enormous: between 100 and 200 billion custom AI and memory chips per year, an initial goal of 100,000 wafer starts per month, and a path toward one million, potentially approaching 70% of TSMC’s total output concentrated in a single U.S. facility. Estimated project cost: roughly $25 billion.

Tesla’s fifth-generation AI chip, AI5, is expected to be among the first products associated with the initiative, with small-batch production beginning in 2026 and larger-scale volume expected in 2027.

To be clear, March 21 almost certainly marks a formal kickoff — a groundbreaking, unveiling, or strategic announcement — not a functioning fab. Facilities of this scale take years to build and commission.

But the signal matters.

Tesla is moving into the same category as Apple, Google, Amazon, and Microsoft: technology companies seeking to control their own silicon destiny.

When the largest AI compute consumers begin owning their own chip supply chains, the semiconductor industry does not merely evolve.

It gets structurally reorganized.

What It All Means

The energy bottleneck that seemed poised to constrain AI is being challenged simultaneously through fusion breakthroughs, private capital surging into next-generation power, nuclear reactivation, and vertical integration across the chip stack.

This is what abundance thinking looks like in practice.

When problems become large enough, urgent enough, and economically important enough, solutions scale to meet them.

The constraint is not permanent.

It never was.

The Supersonic Tsunami: How the System Fits Together

This is what Elon saw so clearly: these are not isolated trends.

They are one integrated, compounding system.

More efficient chips make AI dramatically cheaper. Cheaper inference allows intelligence to be deployed everywhere. That enables agentic systems to run locally inside robots, vehicles, factories, and edge devices.

Fusion and next-generation energy expand power availability. That supports larger training clusters. Larger training clusters produce more capable frontier models. Those models move into embodied systems. Humanoids and autonomous machines begin operating in real environments. Cheap launch and orbital infrastructure extend those systems beyond Earth-based industry.

Each layer amplifies the next.

And the capital is already flowing.

Hundreds of billions in AI infrastructure.
$50 billion data centers.
$10 billion annual revenue streams per gigawatt.
Companies are growing from $1 billion to $14 billion in barely over a year.

This is not a theory.

This is a deployment.

The companies being built now are not competing against 2024 business models.

They are emerging inside what can only be called an abundance economy — one where intelligence becomes cheap, energy becomes more plentiful, labor becomes robotic, and access to advanced capability becomes radically more distributed.

This future is not ten years away.

It is arriving now and will be deployed over the next 12 to 24 months.

What You Need to Do

If You’re an Entrepreneur

Stop designing for 2024 scarcity.

Start designing for 2030 abundance.

Assume intelligence becomes cheap, energy becomes plentiful, and robotic labor becomes scalable. Then ask: what becomes possible in that world that is impossible today?

Your edge will not come from slightly better execution on yesterday’s assumptions.

It will come from imagination applied to tomorrow’s reality.

If You’re an Investor

Own the infrastructure.

AI chips. Energy systems. Launch platforms. Robotics. Data centers. Grid technologies. Semiconductor fabrication.

When an industry deploys trillions of dollars into new infrastructure, that is where generational wealth tends to be created.

Follow where the deepest capital and the smartest operators are already placing their bets.

Position before the inflection becomes obvious.

If You’re a CEO

Assume your industry is about to be stress-tested.

Ask the hard question now: what would our business look like if compute were effectively free, energy abundant, and robotic labor scalable?

Then work backward.

The Fortune 500 is resisting because management often fears automating itself out of relevance. Meanwhile, startups are already deploying AI into legal, marketing, accounting, customer support, HR, and software development.

Do not become the company that resists until it is too late.

And above all, hire for imagination — not just execution.

If You’re a Student

Do not compete with AI.

Collaborate with it.

Use it as a thinking partner, a builder, a tutor, a strategist, and a coding collaborator. Stop orienting your future exclusively around getting hired into an existing system. Start thinking about what you can build.

Your generation will be the first truly AI-native generation. You will create companies and institutions older generations cannot yet fully imagine.

For decades, software talent has been the rate-limiting factor in entrepreneurship. Now one person with AI tools can do work that once required entire teams.

Design your learning around compounding capability, not just your first job.

The Bottom Line

Elon called it a supersonic tsunami.

This past week made clear he was right.

Anthropic is going from $1 billion to $14 billion in 14 months.
OpenAI is moving toward what could become the largest AI IPO in history.
$50 billion data centers generating $10 billion annually.
Trillions flowing into infrastructure.
Neuromorphic chips are making AI dramatically more efficient.
Fusion moving from laboratory promise toward grid relevance.
Humanoids moving from demos into real-world deployment.
Orbital access is becoming routine.

These are not separate headlines.

They are all expressions of the same underlying reality:

Multiple exponential technologies are crossing deployment thresholds simultaneously.

The people who understand that compute, energy, intelligence, embodiment, and orbital access form one compounding system will be the ones who build what comes next.