Investing in the Singularity
Wealth creation is going exponential for those investing in the “Inner Most Loop”… Today I want to show you evidence that can’t be ignored. Before I share and demonstrate the details, let me be clear up front: I’m not a financial advisor and this isn’t financial advice. I’m an engineer and entrepreneur who’s spent 40 years watching exponential technologies blindside entire industries. Following are the numbers I think every investor needs to understand.
On my Moonshots podcast, we keep coming back to what the Moonshot Mates call the innermost loop: the chips, data centers, frontier AI labs, and power companies feeding them. These four sectors are the engine room of the intelligence revolution.
Everything else in the economy is downstream.
In March 2026, I gave a talk to my Abundance Community where I put up a slide showing NVIDIA’s annual revenue hit $216 billion, which was larger than the GDP of 160 countries and territories, which represent more than 70% of the world’s economies. The room of 600 entrepreneurs and investors who are not easily impressed went dead quiet.
I told them that day: “This isn’t a tech cycle. This is the beginning of the largest infrastructure buildout in human history.” I want to show you what happened next.
The One-Year Scoreboard (are shocking)
From May 2025 to May 2026, the S&P 500 returned roughly 31%. That’s great. A rising tide. If you just parked money in an index fund, you did well.
But look at what happened inside the sectors driving the Singularity… Chips, Data Centers and Energy.
Semiconductors (the brains of AI) one-year period, May 2025 - May 2026:
Micron (MU): +770%
Intel (INTC): +483%
AMD: +343%
Taiwan Semiconductor (TSM): +133%
Broadcom (AVGO): +107%
NVIDIA (NVDA): +85%
Over the past year, these six AI chip companies gained an average of roughly 320%, about 10× the S&P 500’s roughly 31% return.
(Intel’s number looks shocking, and it is. Intel was left for dead in mid-2024, with its stock trading below $20, but then executed a turnaround driven by their foundry business and the CHIPS Act. That’s a separate story worth an entire newsletter.)
Now let’s look at the same one-year period, May 2025 - May 2026, for three of the leading Data Center Infrastructure companies, and two of the Power Companies Feeding the Machine:
1. Vertiv (VRT): ~+256% – They build the cooling, power distribution, and racks that keep AI running.
2. GE Vernova (GEV): ~+158% – Turbines, grid infrastructure, nuclear services.
3. Bloom Energy (BE): ~+1,647% – Fuel cells for onsite data-center power.
4. Oklo (OKLO): ~+178%% – Advanced nuclear for data centers.
5. Fluence Energy (FLNC): ~+200% – Battery-storage and grid-scale energy systems.
6. Talen Energy (TLN): ~+73% – Nuclear power for AWS.
Now compare that to traditional sectors over the same one-year window:
S&P 500 (SPY)*: ~+32%
Financials (XLF): ~+5%
Healthcare (XLV): ~+8%
Consumer Staples (XLP): ~+6%
Industrials (XLI): ~+29%
* Note: Roughly one-third of the entire S&P 500 gain came from just five chip stocks: NVIDIA, Broadcom, AMD, Micron, and Intel.
The six “Energy & Infrastructure stocks” average return is about +419%, versus roughly +32% for SPY. Figures are rounded from reported 12-month total returns through early May 2026.
The Hyperscalers Are Pulling Away
The public companies building and operating AI at scale (i.e., the hyperscalers) have also diverged dramatically from the broader market over the past two years.
In May 2024, Alphabet (Google) had a market cap of roughly $2.1 trillion. Today it sits at $4.81 trillion, a 129% gain, making it the world’s second most valuable company. Google’s AI investments through DeepMind, Gemini, and its cloud platform have been rewarded massively.
Amazon was about $1.9 trillion in May 2024. Today: $2.93 trillion (+54%), driven heavily by AWS’s AI infrastructure dominance. Meta was about $1.2 trillion. And today it’s $1.57 trillion (+31%), powered by its open-source Llama models and massive AI compute buildout.
Even Microsoft, which had already run up significantly before our two-year window, is at $3.07 trillion, roughly where it was in May 2024. But here’s the thing about Microsoft: through its partnership with OpenAI and Copilot integration across every product, it may be the most AI-leveraged large company on the planet. The market is still trying to price in how much of Microsoft’s future revenue is AI-driven.
Collectively, these four hyperscalers are now worth over $12.4 trillion, and they’re projected to spend approximately $750 billion on AI capex in 2026 alone. That $750 billion is more than the GDP of Switzerland. Flowing into chips, data centers, and power.
In a single year.
Frontier Labs: The Fastest-Growing “Companies” in History
But here’s where the numbers get truly wild. The private frontier AI labs, the companies actually building the foundational models, are experiencing valuation growth that has no historical precedent.
OpenAI: Valued at approximately $80 billion in early 2024. On March 31, 2026, Bloomberg reported they closed a $122 billion funding round (the largest private raise ever) at a post-money valuation of $852 billion. Revenue went from $2 billion (2023) to $6 billion (2024) to $20 billion (2025), confirmed by CFO Sarah Friar in January 2026. They now have 910 million weekly active users. They’re planning an IPO that could value them north of $1 trillion.
That’s a 10x valuation increase in two years. For context, it took Google 6 years to go from $80 billion to $800 billion. OpenAI did it in 24 months.
Anthropic: Valued at roughly $18 billion in early 2024. In February 2026, they closed a $30 billion round at $380 billion. Then it kept going. In late April, TechCrunch reported Anthropic had received preemptive offers to raise $50 billion at a valuation between $850 billion and $900 billion. And as of early May, Business Insider reported that Anthropic’s secondary market valuation on Forge Global had reached $1 trillion, surpassing OpenAI. A 55x increase from where they started in early 2024. Forbes expects an IPO possibly by October 2026.
xAI (Elon Musk’s lab): In January 2026, xAI closed a $20 billion Series E at a $230 billion valuation, according to Sacra and Tracxn. Total funding raised: $45 billion. Revenue grew from roughly $100 million in 2024 to $3.8 billion annualized by the end of 2025, 38x year-over-year growth.
Google DeepMind: While not separately valued, DeepMind’s work on Gemini and AlphaFold has been a central driver of Alphabet’s surge from $2.1T to $4.81T. Sundar Pichai has called AI “the most profound technology” Google has ever worked on. And the market agrees: Alphabet’s 129% two-year gain is essentially a DeepMind premium.
Add these up and the picture is clear: frontier AI lab valuations have collectively grown by over $2 trillion in two years. And funding to foundational AI startups in Q1 2026 alone doubled the total raised in all of 2025, per Crunchbase.
So, What’s Driving All of This?
Three numbers tell the story of the underlying demand.
$1 trillion. Global semiconductor sales are on track to hit $1 trillion in 2026, per the Semiconductor Industry Association. Q1 2026 alone came in at nearly $300 billion, with March revenue at $99.5 billion in a single month. The entire year of 2023 was about $527 billion. The chip industry is roughly doubling in three years.
$750 billion. As I mentioned above, that’s the projected AI capital expenditure from the big hyperscalers in 2026. CreditSights recently raised their estimate from $650 billion. Every dollar flows into chips, power systems, cooling infrastructure, and physical data center construction.
$2.02 trillion. That’s where analysts project the AI data center market will be by 2032, growing at 27.5% compounded annually from $471 billion in 2026. The smartphone market took about 15 years to reach similar scale. AI infrastructure is on pace to get there in six.
Elon’s GDP Prediction
On Christmas Day 2025, Elon Musk posted on X: “Double-digit growth is coming within 12 to 18 months.” He added that if applied intelligence is treated as a proxy for economic growth, “triple-digit is possible in about five years.”
Most economists dismissed this. Consensus GDP forecasts remain in the 2-3% range. But look at what the market is actually doing. Not what economists predict, but where capital is flowing. $750 billion in hyperscaler capex. $1 trillion in global chip sales. Frontier labs doubling their funding every quarter. The financial markets are pricing in something far closer to Elon’s view than to the consensus forecast.
Whether you call it double-digit GDP or not, the economic engine driving AI infrastructure is already growing at rates that would be double-digit if measured as its own economy. The AI infrastructure sector alone is adding hundreds of billions in new economic activity each year, and accelerating.
The Innermost Loop… Models, Chips, Data Centers & Power
Here’s how the loop works:
Frontier labs build the models. They can’t build models without chips: NVIDIA GPUs, AMD accelerators, Broadcom custom ASICs, Micron’s HBM memory. Those chips need to live in data centers, built and cooled by Vertiv, hosted by Equinix and Digital Realty.
And those data centers need power, which is why Constellation and Vistra are bringing nuclear plants back online, GE Vernova is selling gas turbines as fast as they can manufacture them, and $156 billion worth of data center projects have been blocked or stalled by local opposition, according to CNBC and Data Center Watch.
Each layer feeds the next. Frontier lab demand drives chip orders. Chip orders drive data center construction. Data center construction drives power demand. Power demand is now the single biggest bottleneck.
The entire loop is accelerating simultaneously. That’s what makes this moment different from the dot-com era or the smartphone buildout. Back then, you had one supply chain under pressure. Today there are four interlocking supply chains all hitting escape velocity at the same time.
Accelerating Towards the Singularity… Where This Goes Next
My friend Ray Kurzweil predicted the Singularity decades ago, the moment when computation becomes abundant enough to bootstrap intelligence itself. We’re inside that prediction now. The market data over the last two years is much more than a bull run.
It’s the economy re-pricing itself around a new center of gravity.
The companies building the physical infrastructure of AI (the chips, the data centers, the power plants) have returned 3x to 6x what traditional sectors have delivered over the same period. The frontier labs building the models have seen 10x to 55x valuation increases. And the hyperscalers funding the entire buildout are spending more on AI infrastructure in a single year than many countries produce in GDP.
Global chip sales are still growing at 25%+ annually. AI capex is accelerating. We’re still early.
- Peter
More From Peter
If you’ve enjoyed Metatrends, here are more ways to stay connected:
The trailing returns are undeniable and the data presentation is effective. But the most important analytical question this piece raises is one it doesnt quite ask, and the answer determines wether these numbers represent the beginning of a multi-decade buildout or the peak of the most concentrated capital cycle since fibre optics in 2000.
Every number in this piece is a trailing return. Micron +770%. Bloom Energy +1,647%. Intel +483%. Extraordinary. But trailing returns tell you what already happened. The investment question is always about whats priced in versus whats still ahead. And the way to test that is to ask: what has to be true about the future for these valuations to be sustained?
For Nvidia at $216B revenue, the market is pricing continued GPU demand at current or higher volumes for at least 3-5 more years. Thats a reasonable assumption if the scaling thesis holds. But Jensen Huang himself said the multi-gigawatt training campus era is "largely behind us" and the future is distributed inference. If inference requires fewer GPUs per unit of compute than training does, Nvidias revenue trajectory flattens even if total AI compute demand keeps growing. The market hasnt priced the architectural shift.
Bloom Energy at +1,647% is the number I keep coming back to because it reveals exactly how much specificity is embedded in these returns. A 16x gain on a fuel cell company requires the market to believe that onsite fuel cells will power the next generation of data centres rather than grid electricity, renewables, or nuclear. Thats a very precise technological bet. If the distributed inference thesis proves correct and compute moves to the edge and into existing commercial buildings, the massive centralised data centre buildout that justifies Blooms valuation loses its demand base. The return was real. The question is wether the thesis underneath it survives the next architectural transition.
The one-third of S&P gains coming from five chip stocks is the data point that should concern every passive investor reading this piece. The S&P 500 returned 31%. Roughly 10 points of that came from five companies. The other 495 companies in the index generated a collective return of about 21%. The "rising tide" wasnt a tide. It was a wave concentrated in a very small part of the ocean, and every index fund holder in America is riding that concentration wether they chose to or not.
The innermost loop thesis is correct about where the value creation is occuring. The Perez framework is the historical counterweight: every major infrastructure buildout in history has produced extraordinary returns for early investors followed by a brutal shakeout that destroyed capital for late ones. Railroads, electricity, automobiles, fibre optics, each one followed the same pattern. Extraordinary returns during the installation phase. Crash during the overcapacity phase. Sustained but lower returns during the deployment phase that follows. The question for anyone reading this piece and deciding wether to allocate capital to the innermost loop today is which phase we're currently in, and the honest answer is that the data supports both "still early" and "approaching the transition" simultanously.
The electricity demand required by the global expansion of data centers needed to power AI growth is enormous. Best investments seem to be in the areas of batteries, renewables and nuclear energy. Betting on the infrastructure needed in the scene to support AI/data center expansion is a smart investment. Like betting on companies that built roads, bridges etc when automobiles came on the scene. Didn’t matter which auto company won , roads and bridges etc would be needed.