AI Is Big, But the Opportunity Is Bigger

• Past technology infrastructure cycles have peaked at 2% to 5% of GDP.

• The AI infrastructure TAM is expected to triple by 2030.

• Nvidia’s forward P/E is far below Cisco’s dotcom peak.

Every cycle has its doubters, but history has a habit of rewarding the steady hand.

There’s an old line in investing that’s grown truer to me with time: “Markets aren’t about timing — they’re about time.” Wealth doesn’t come from guessing the next swing. It comes from building a plan, sticking to it, and letting compounding do the heavy lifting. Diversification, consistent contributions, and reinvested dividends are the quiet engines of long‑term growth.

Take the S&P 500. Since 1928, it’s delivered roughly 9.7% a year with dividends reinvested. That’s enough to double your money about every 7.5 years, through wars, recessions, inflation spikes, and political drama. A simple $100 invested at the end of 1928 would be worth around $590,000 today. That’s the power of staying put…

The last few months have been a reminder of why patience matters. After falling nearly 8% in February and March, the market snapped back sharply. With the U.S. and Iran reaching a cease‑fire agreement and first‑quarter earnings beating expectations, the S&P 500 has surged 16% off the lows.

But today I want to focus on a different source of anxiety: the growing chorus calling tech a bubble. The current technology weighting has reached 39.6% of the S&P 500, above its 34.8% at the dot‑com peak. That alone was enough to spark “top‑calling season.”

I understand the instinct. The late 1990s left a scar. And there’s always someone eager to make their career by calling the exact top. If they nail it, they’ll be quoted for decades. But the data doesn’t support the bubble narrative. The S&P 500 still has room for a durable, grinding advance.

But don’t take my word for it, let’s look at what the data’s telling us…

AI Infrastructure Spending

If you’re hunting for signs of an AI bubble, capex is the first stop. Critics point to the sheer scale of investment.

AI‑related data‑center spending is expected to reach $650 billion or more this year. Alphabet, Amazon, Meta, and Microsoft alone account for the bulk of that — well above the early 2026 Wall Street estimate of $550 billion.

But history puts this in perspective…

U.S. nominal GDP in 2026 is projected at $32.4 trillion. Against that backdrop, roughly $650 billion in AI capex is only about 2% of GDP. Historically, major infrastructure cycles have peaked between 2% and 5% of economic output. For comparison, last year’s U.S. defense budget was 3% of GDP.

In other words: this isn’t unprecedented. It’s on the low end.

Total Addressable Market Growth

Next, let’s look at growth potential. The total addressable market (“TAM”) is the biggest possible pool of buyers. Companies use it to gauge how much revenue they could capture if everyone bought in.

Few executives have been as transparent about AI infrastructure demand as Lisa Su, CEO of AMD. Her projections tell a clear story:

• In 2023, she estimated $400 billion in AI infrastructure by 2027.

• By early 2025, that rose to $500 billion by 2028.

• This year, she said TAM could exceed $1 trillion by 2030.…

She also expects global AI users to grow from 1 billion to 5 billion over the next five years.

Why the surge? Because AI is becoming embedded in everything — healthcare, transportation, enterprise software, personal devices. Agentic AI requires enormous compute power, and consumer hardware is racing to keep up.

Goldman Sachs now expects total annual capex to reach $1.6 trillion by 2031, up from $765 billion in 2026. As AI adoption spreads, the TAM expands with it.

Infrastructure Darlings: Now vs. Then

Skeptics love the Nvidia‑Cisco analogy. Both were the backbone of their eras… Nvidia for AI, Cisco for the early internet.

The argument: Cisco overbuilt, demand fell short, and the stock never recovered. Nvidia could face the same fate.

But the valuation data tells a different story. Look at the forward price-to-earnings ratio for Nvidia since 2021 compared to Cisco from 1995-2000…

Nvidia’s forward P/E in 2021–22 looks similar to Cisco’s in 1995–96. But from there, the paths diverge:

• Cisco’s multiple exploded higher into the bubble.

• Nvidia’s multiple has compressed even as the stock has rallied.

That’s not what mania looks like. That’s what earnings power looks like.

The Nasdaq Composite

Finally, let’s compare the Nasdaq’s trajectory. I used 2023 as the starting point for today (post‑ChatGPT) and 1995 for the dot‑com era (post‑Mosaic browser).…

Since early 2023, the Nasdaq is up ~145%.

• From 1995 to the same point in 1997, it gained 152%.

• But by March 2000, the dot‑com Nasdaq had surged nearly 570%.

• To match that, today’s Nasdaq would need to triple from here.

We’re nowhere near that kind of euphoria.

Bringing It All Together

Yes, today’s market has echoes of the dot-com boom, like tech’s heavy weighting in the S&P 500. And yes, some people want to call the top.

But history shows that new technologies drive growth and investment cycles forward. That’s what happened before, and it’s happening again now. Based on precedent, we’re not near a peak. That backdrop supports a steady long-term rally in the S&P 500.

Five Stories Moving the Market:

The Bank of Japan needs to raise interest rates every few months and should quicken the pace in the face of rising inflationary pressures, according to board member Naoki Tamura – Bloomberg. (Why you should care – Tamura, whose term expires next year, has been one of the more hawkish BOJ members)

Micron Technology forecast quarterly earnings above Wall Street ‌estimates; the company expects fourth-quarter revenue of $50 billion, plus or minus $1 billion, compared with analysts' average estimate of $43.58 billion and earnings of $31 versus the expectation for $25.84 – Reuters. (Why you should care – the guidance speaks to the continued strength for AI-related semiconductors)

All the biggest U.S. banks cleared the Federal Reserve’s annual stress test, setting the stage for lenders to boost buybacks and dividends; the exam aims to gauge how Wall Street lenders would fare under a hypothetical shock to the financial system – WSJ. (Why you should care – the Fed said capital remained above minimum requirements, despite absorbing more than $798 billion in losses under the hypothetical stress test scenario)

Qualcomm said that Microsoft and Meta Platforms will use its new AI chips and that it will make custom chips for two other unnamed "hyperscalers"; the comments were made during an investor presentation to showcase its move into supplying AI chips ​for data centers – Reuters. (Why you should care – management said it expects AI chip sales to contribute to overall revenue before the end of this year)

The Bank of Japan is increasingly highlighting the economic benefits of the global artificial intelligence boom, viewing strong demand for AI-related products as a buffer against the drag from higher energy prices linked to the Middle East conflict – Bloomberg. (Why you should care – a rebound in economic growth could place pressure on policymakers to raise interest rates more)

Economic Calendar:

Earnings: CMC, DRI, MKC, WGO

Germany – GfK Consumer Climate for July (2 a.m.)

ECB Economic bulletin (4 a.m.)

ECB’s Lane (Chief Economist) Speaks (6 a.m.)

U.S. – Chicago Fed National Activity Index for May (8:30 a.m.)

U.S. – GDP (Third Take) for Q1 (8:30 a.m.)

U.S. - Initial Jobless Claims (8:30 a.m.)

U.S. - Continuing Claims (8:30 a.m.)

U.S. – PCE for May (8:30 a.m.)

U.S. – Personal Income and Spending for May (8:30 a.m.)

U.S. – Kansas City Fed Manufacturing Index for June (11 a.m.)

Treasury Auctions $44 Billion in 7-Year Note (1 p.m.)

Fed’s Williams (New York, Voter) Speaks (3:40 p.m.)

Fed's Balance Sheet Update (4:30 p.m.)

Fed’s Goolsbee (Chicago, Non-voter) Speaks (6:30 p.m.)

Japan – Tokyo CPI for June (7:30 p.m.)