I Witnessed the AI Compute Supercycle in Taiwan. Here's What Nobody Is Telling You About the 2026 Boom.

You want to know what it looks like when history bends? Let me paint you a picture.

Picture Taipei, right now. The news isn't about politics. It's not about the weather. It's about a man in a leather jacket: Jensen Huang, the CEO of Nvidia. People are literally lining up in the streets, clutching phones and notebooks, hoping to catch a glimpse of him. They call him the "Godfather of AI."

And it's not just hero worship. It's something else. Something more primal. It's FOMO.

People who have never bought a single stock in their lives are now talking about quitting their jobs to trade full-time. Why? Because their friends, neighbors, and cousins have seen their "unrealized profits" on AI stocks grow larger than their annual salaries. They're watching the SOXX semiconductor ETF climb 25% in a single month, with seemingly no end in sight.

The AI compute supercycle isn't coming. It's here. And it's rewriting the rules of the global economy in real-time.

This isn't the crypto boom of 2021. This isn't the .com bubble. This is something far more fundamental. We are witnessing the physical construction of the world's new digital brain, and a handful of companies are supplying the neurons.

In this post, we're going to tear apart the "Compute Components Super Saiyan Cycle," explaining exactly why you're seeing shortages, insane valuations, and the biggest IPO in history. More importantly, I'm going to show you how to navigate this chaos as an investor, without losing your shirt.

The Scene in Taiwan: A FOMO-Fueled Frenzy

Let's start with the picture I just described because context is everything.

The original article that sparked this post noted that in Taiwan, people are "leaving their jobs to focus on trading." Think about that for a second. People are abandoning the security of a steady paycheck for the volatility of day trading. That's not an economic indicator. That's a fever dream.

Jensen Huang's recent visit has become a national event. His face is on every screen. His company's $150 billion-a-year investment pledge has turned Taiwan from a manufacturing hub into the self-proclaimed "epicenter of the AI revolution."

But here's what the headlines miss. The frenzy isn't just about hype. It's happening because people can see the pipeline with their own eyes. They see the factories running 24/7. They see the orders for Nvidia's GPUs stretching out for nearly a year. They know that a new "Grace CPU" is being deployed specifically for Meta's personal AI agents.

This isn't speculation about a digital future. It's a direct observation of physical reality. The world is building something enormous, and Taiwan is the hardware store.

The "Super Saiyan Cycle": Why This AI Boom Is Different

Every tech cycle has a story. The dot-com boom was about connecting everyone to the internet. The mobile boom was about putting a computer in everyone's pocket. This cycle is different. And honestly? The phrase "Super Saiyan Cycle" is the best metaphor I've heard yet.

In the anime Dragon Ball Z, a Saiyan gets exponentially stronger every time they recover from a near-fatal battle. That's exactly what's happening to the AI hardware market. Every time we think we've hit a limit, a new bottleneck emerges, and then a new solution emerges to break through it, only to create the next bottleneck.

From Software to Silicon: The Physical AI Pivot

For the last two years, the AI story was about software, ChatGPT, Claude, and the chatbots. That era is ending.

We've now entered the "physical AI" phase. The focus has flipped from training models in the cloud to running them everywhere else: in your car, on your phone, in a factory robot. This is what analysts call the "inference flip."

And here's the kicker. Inference requires a vastly different, and vastly more diverse, set of hardware than training. It means we need custom AI processors, low-latency memory, and advanced packaging solutions for everything from smart fridges to autonomous drones.

The old era was about one or two supercomputers. The new era is about putting a million smart chips into the physical world.

The Three Chokepoints: Chips, Memory, and Power

So why the shortage? It's simple: Demand is exploding across three critical fronts, and supply isn't just tight, it's a disaster.

The Chip Chokepoint: It starts with the GPU. Nvidia's H100 and B200 chips are the gold standard, and orders have ballooned to a staggering $1 trillion through 2027. Lead times are stretching to nearly a year. You can't just order a high-end AI chip and expect it next week.
The Memory Chokepoint (HBM): This is the silent killer. You can have the fastest chip in the world, but if it can't access data quickly, it's useless. High-Bandwidth Memory (HBM) is the "short-term memory" of an AI accelerator. In 2026, the supply gap for HBM is a mind-blowing 50% to 60%. This single component could account for 30% of all AI spending this year, up from just 8% two years ago.
The Power Chokepoint: All these chips need electricity, and a lot of it. We're not just talking about a few more power plants. We're talking about a complete overhaul of national grids. An AI data center consumes up to three times more copper for power distribution than a traditional cloud facility. We are quite literally running out of the metal needed to plug everything in.

Who's Feeling the Squeeze? The AI Supply Chain Exposed

This isn't just an Nvidia problem. It's a problem for everyone. Here's how the "compute shortage" ripples down:

Hyperscalers (Amazon, Google, Microsoft): They're throwing money at the problem. Their combined capital expenditure is expected to hit $602 billion in 2026. They're locking in supply contracts years in advance and even designing their own custom chips.
The Rest of the World: That's the scary part. If you're not a trillion-dollar tech giant, you're being pushed to the back of the line. Startups, research labs, and even governments are finding it nearly impossible to secure the compute they need.

The Headline Trio: Market Movers You Can't Ignore

All of this theoretical talk is interesting, but the market is where it gets real. Three stories are currently dominating the financial news, and they are three perfect expressions of the AI compute supercycle.

Jensen Huang's $150B Love Letter to Taiwan

Let's start with the "Godfather of AI" himself. During his COMPUTEX keynote in Taipei, Jensen Huang made a stunning announcement: Nvidia plans to invest roughly $150 billion a year in Taiwan.

He called the country the "epicenter of the AI revolution." He unveiled plans for a new, "transparent" headquarters with glass curtain walls, set to begin construction in late 2026.

This isn't charity. It's a strategic necessity. Taiwan is where the vast majority of the world's most advanced chips are manufactured, primarily by TSMC. Nvidia is effectively building a fortress around its most critical supply chain.

For investors, this is a massive vote of confidence in the entire Taiwan semiconductor ecosystem. As Jensen goes, so goes the AI hardware world.

SpaceX's IPO: The Largest Debut in History?

If there was any doubt that we're in a supercycle, the SpaceX IPO should erase it.

After years of speculation, Elon Musk's rocket company is finally going public, targeting a Nasdaq listing as early as June 12, 2026, under the ticker "SPCX."

And the numbers are astronomical. SpaceX is looking to raise up to $75 billion at a valuation of roughly $1.75 trillion. That would make it the largest IPO in the history of the world, instantly becoming one of the most valuable public companies on the planet.

How is this relevant to an AI compute cycle? Because SpaceX is not just a space company. It's an AI company. Its Starship rockets, its Starlink satellites, and its mission control systems are all powered by advanced AI. More directly, the company has signed a $15 billion annual revenue deal with Anthropic to provide compute for its AI models. SpaceX needs the chips to fly its rockets, and it's building the infrastructure to supply them to others.

Anthropic's Series H: A $965B Bet on Safe AGI

Finally, we have Anthropic. In late May 2026, the AI company behind the Claude model closed a massive $65 billion Series H funding round, led by the likes of Sequoia Capital and Altimeter Capital. The round valued Anthropic at a jaw-dropping $965 billion.

To put that in perspective, that valuation now surpasses its chief rival, OpenAI, which is valued at $852 billion. Anthropic went from a $380 billion valuation in February to nearly $1 trillion in just three months.

Why the insane valuation? Because investors are placing a bet on two things: 1) that "safe" and "aligned" AI will win the enterprise market, and 2) that Anthropic knows how to scale. The company's run-rate revenue had already crossed $47 billion earlier in May, and it expects 130% year-on-year growth.

Anthropic is going to take that $65 billion and spend a huge chunk of it on exactly what we've been talking about: compute, compute, and more compute. They need the chips to train their next-gen models and run their inference engines. They are a direct beneficiary and a direct driver of the supercycle.

How to Ride the AI Supercycle (Without Getting Crushed)

Okay, enough about the macro. You want to know what to do.

First, a word of caution. The energy in Taiwan is the same energy we saw in 1999. That doesn't mean this is a bubble that's about to pop. But it does mean we're in a period of extreme euphoria, and extreme euphoria is followed by extreme volatility.

ETFs vs. Single Stocks: A Risk Breakdown

You have two ways to play this.

Single Stocks (Nvidia, TSMC, AMD): The upside is massive. If Nvidia continues to dominate, a direct investment could 5x or 10x over the next decade. But the downside is equally massive. One earnings miss, one piece of antitrust legislation, or one successful competitor could wipe out a huge chunk of your investment overnight.
Semiconductor ETFs (SOXX, SMH): This is the boring, smarter play for most people. An ETF like SOXX tracks a broad index of 30 U.S. chip stocks. It won't make you a millionaire overnight. But it will give you exposure to the entire value chain. When one company stumbles, another will rise. It's the "own the pickaxes in a gold rush" strategy. And in this case, the pickaxes are made of silicon.

A Simple 3-Step Plan for Smart Exposure

If I were building a portfolio today to capture the AI compute supercycle, this is how I'd do it.

Step 1: Establish Your Core (50-70%)
- Put the majority of your AI investment into a broad semiconductor ETF like SOXX or SMH. This is your foundation. It will capture the overall trend without exposing you to company-specific risk.
Step 2: Pick Your Winners (20-30%)
- If you have a strong thesis about a specific company, allocate a smaller portion to single stocks. Nvidia (NVDA) is the obvious pick. TSM (Taiwan Semiconductor) is the quiet giant behind everyone. AMD is the plucky underdog with a lot to prove.
Step 3: Watch the Infrastructure (10%)
- Look beyond the chipmakers. Companies that provide the infrastructure for these chips are critical. This includes companies like ASML (makes the machines that make the chips) and Freeport-McMoRan (FCX) (a major copper miner). Data centers run on copper, and they're about to run on a lot of it.

And remember: Never invest money you can't afford to lose. This is a supercycle, not a straight line. There will be crashes, pullbacks, and terrifying moments. Zoom out. Stay calm. And stick to the plan.

The Final Word: Waking Up to a New Economic Reality

When people start lining up in the streets for a CEO's autograph, it's easy to dismiss it all as a cult or a mania.

But what's happening in the world of AI hardware is far more profound. We are watching a complete rewiring of the global economy. The AI compute supercycle is not a trend. It's the new foundation.

The question isn't whether you'll be affected by it. You already are, every time you use a search engine, scroll through a social feed, or ask a virtual assistant a question. The only question is whether you'll be an informed observer or an active participant.

Jensen Huang is betting $150 billion a year on Taiwan. Elon Musk is betting the largest IPO in history on a rocket company that needs AI to fly. The world's top VCs just poured $65 billion into a startup to make it safer.

The smart money has already picked its side. The only question left is: Have you?

BBD Investors

Search This Blog