Everyone Is Betting on AI — But Almost No One Asks: Where Are You Standing?The AI boom is real. The harder question is where value, risk, and leverage accumulate inside the stack.
Position matters more than opinion.
Over the past few weeks, I have been thinking about AI through a different question:
Why can the same AI cycle feel so powerful in markets, yet so uncertain in daily life?
My own view is more optimistic than pessimistic.
It is hard to dismiss the physical scale already underway: AI-related capital expenditure, semiconductor demand, data centre construction, networking equipment, power infrastructure, and software experimentation are no longer just narratives. They are orders, capex plans, fab loading, land, electricity, and supply chains.
But optimism needs a map.
This is not a Taiwan-versus-America story.
U.S. equity markets have also been lifted by the AI cycle, especially through mega-cap technology companies, cloud platforms, and AI infrastructure spending. Taiwan’s current boom is deeply connected to that same demand — it is one of the most important hardware and supply-chain expressions of the cycle.
The contrast is not between countries.
It is between positions inside the same system.
Platforms, chips, data centres, power, capital markets, labour markets, and households do not experience the AI cycle in the same way.
That is why the same cycle can feel euphoric to asset owners, strategic suppliers, and people close to the bottlenecks — while still feeling uncertain to workers, families, and companies whose roles are being repriced.
Both feelings can be true.
They are simply coming from different layers of the stack.
That is what pushed me to write this essay.
Not to argue that AI is just hype.
I think this is a real structural investment cycle, with meaningful productivity potential. But it also carries valuation risk, execution risk, and distribution risk.
The real question is no longer:
Will AI change the world?
A more useful question is:
When it does, where exactly do we stand?
This series is an attempt to answer that question from the bottom up — from process technology, semiconductors, photonics, and infrastructure, all the way to industrial structure and capital allocation.
It is not about predicting stock prices.
It is about understanding what must be true before risk and opportunity can be properly aligned.
1. The AI Boom Is Real. The Distribution Is Not Automatic.
The AI boom is increasingly difficult to dismiss as pure hype.
The money is real. The orders are real. The data centres are real. The power constraints are real. The semiconductor bottlenecks are real.
But that does not mean every part of the economy benefits in the same way, at the same time, or with the same intensity.
This is where the current cycle becomes difficult to describe with simple labels.
Calling it a “bubble” misses the physical investment already underway.
Calling it a “revolution” can also be too vague, because revolutions do not distribute gains evenly.
A more useful framing is this:
AI is a real structural investment cycle with valuation, execution, and distribution risks.
That distinction matters.
In the United States, AI-related investment has become an important part of the macro discussion. The St. Louis Fed has tracked contributions from information processing equipment, software, R&D, and data centre investment to real GDP growth, showing that several AI-adjacent investment categories were meaningfully elevated in parts of 2025. [2]
At the same time, the accounting is complicated. Some AI infrastructure investment flows through imported chips and equipment, so the domestic GDP effect can be harder to measure than the market narrative suggests. A Federal Reserve paper on data centre investment notes that GDP contributions depend not only on the level of investment, but also on import shares and the pace of new project inflows. [3]
This is exactly why the story feels both real and confusing.
AI can lift capital expenditure, equity markets, exports, and strategic investment — while still leaving many people unsure whether their own position is improving.
That does not make the boom fake.
It makes the position more important.
2. Markets Are Booming. The Gains Are Uneven.
One of the more uncomfortable realities in this AI cycle is this:
Markets, exports, and capex can surge without translating one-for-one into broad-based job security, wage bargaining power, or career mobility.
This is not a contradiction.
It is a structural feature of the current cycle.
In Taiwan, this is visible in the concentration of growth around AI hardware, semiconductors, advanced manufacturing, and the supply chains around them.
Taiwan’s official statistics show how strong this cycle has become. DGBAS reported that Taiwan’s real GDP grew 8.68% in 2025 and projected 7.71% growth in 2026, citing AI-driven external demand and expanding semiconductor and ICT-related investment. [4]
Taiwan’s Ministry of Finance also reported record-high exports in 2025. Information, communication and audio-video products reached US$251.2 billion, while electronic parts reached US$222.9 billion; together they accounted for about 74% of total exports. The same report explicitly linked integrated circuit export growth to demand for AI and high-performance computing. [5]
That is a remarkable industrial position.
But it is also a concentrated one.
A small number of large companies can move an entire index, while many traditional industries and service sectors recover more slowly. The supply chain can be booming even if the average worker does not feel an equivalent improvement in bargaining power.
In the United States, the pattern looks different but rhymes.
AI platforms, cloud capex, and mega-cap technology companies can lift equity markets and business investment, while many households still worry about jobs, inflation, housing, insurance, and long-term security.
The AI economy is capital-intensive, skill-intensive, and position-sensitive.
It can create enormous output, market value, and strategic leverage without distributing the benefits evenly across every layer of labor, households, and firms.
That is why the better question is not simply whether AI creates growth.
It clearly can.
The better question is:
Who is close enough to the value creation layer to feel that growth directly?
3. “Betting on AI” Is Not a Strategy.
“Betting on AI” is often used as if it were a decision.
In reality, without specifying where, it is mostly sentiment.
Most people think they are choosing a direction. In reality, they often stay at the most abstract layer.
The AI industrial system can roughly be decomposed into several layers:
- Models and Applications
Cloud services, agents, enterprise software, automation tools, content systems, and workflow products.
- Compute and Infrastructure
GPUs, ASICs, networking, switches, servers, data centres, cooling, and power delivery.
- Semiconductors and Photonics
CPO, silicon photonics, memory, advanced packaging, wafer fabrication, testing, and process integration.
- Energy and Land
Power generation, grid access, thermal management, water, site selection, permits, and physical infrastructure.
- Talent and Institutions
Education systems, research ecosystems, industrial clusters, governance, regulation, and long-term human capital.
Saying “AI is promising” without anchoring that belief to one of these layers has little operational meaning.
The real questions are:
-
Which layer do you want to commit to over the long term?
-
What is the return mechanism in that layer — equity, cash flow, technical IP, capacity position, or career capital?
-
What capabilities does that layer require — process know-how, patents, supply chain access, capital discipline, or talent density?
Take compute infrastructure, especially CPO and silicon photonics, as an example.
This is not a one-year theme. It is a multi-decade supply chain.
Upstream, there are process and material questions: SiN waveguides, Ge integration, Cu-Cu hybrid bonding, and the equipment and IP required to make these technologies manufacturable.
Midstream, there is fabrication reality: BEOL thermal budgets, yield learning, wafer-level variation, testing feedback, and the ability to move from demonstration to repeatable production.
Downstream, there is system integration: modules, packaging, optical-electrical interfaces, reliability, cost, and commercial adoption.
If you cannot identify who controls the technology, who owns the capacity, and who accumulates learning over time, then “bullish on AI infrastructure” becomes indistinguishable from chasing noise.
4. For Investors: Filter by Technology Before Valuation.
From an investment perspective, AI often feels like a game of speed.
That instinct is dangerous.
It encourages people to skip the most important step: filtering by technical and production reality before discussing valuation.
Take CPO again.
Instead of asking only, “When will CPO scale?”, a more useful set of questions would be:
-
Under BEOL thermal constraints, what is the loss profile of SiN waveguides? What does the AFM roughness distribution look like at wafer level?
-
Is Cu-Cu hybrid bonding done in-house or outsourced? How is Cu recess variation controlled across the wafer?
-
Are Ge photodetectors measured under real integration thermal budgets? What are the trade-offs between dark current, responsivity, and bandwidth under production conditions?
These are questions that most market narratives never even attempt to answer.
But they matter.
Companies that can discuss these problems clearly, even if their numbers are not yet perfect, are at least engaging with reality. Companies that can only talk about vision, total addressable market, and future demand operate under a fundamentally different risk structure.
TSMC’s Q1 2026 transcript is a useful reminder of how physical this cycle is. The company reported that HPC accounted for 61% of first-quarter revenue, discussed strong demand from cloud service providers, and raised confidence that full-year 2026 revenue would grow by more than 30% in U.S. dollar terms. It also described capacity expansion in advanced nodes to meet AI-related demand. [6]
That kind of detail is important because it connects the AI narrative to specific capacity, process nodes, customer demand, and capital allocation.
The point is not that every investor must become a process engineer.
The point is that in capital-intensive, technology-driven industries, technology and capacity are themselves forms of capital allocation.
If this layer is misunderstood, valuation models become misaligned by definition.
5. For Companies: Slides Are Not Strategy.
For management teams, one of the biggest risks in the coming years is treating AI and infrastructure as themes to attach to, rather than constraints to work within.
The real questions are less about presentation language and more about boundaries:
-
Where is our actual technical foundation — models, systems, chip design, fabrication, packaging, or supply chain execution?
-
Within that layer, what are we building — core IP, durable capacity, customer access, or talent pipelines?
-
Are we a replaceable component inside someone else’s ecosystem, or do we hold leverage at a critical node?
The answers may not move the stock price in the short term.
But over five to ten years, they determine whether a company becomes part of the infrastructure or remains a disposable piece of it.
Strategy does not need to be grand.
But if strategy is only incremental adjustment around trending themes, the outcome is often predictable: helping someone else amortise capex while accumulating very little long-term asset value of your own.
That is not positioning.
That is motion.
6. For Individuals and Families: Choose a Line You Can Compound.
At the individual level, the structure is similar.
Short-term tactical decisions are always available. People can trade the cycle, follow headlines, rotate between themes, or react to market sentiment.
But long-term positioning comes down to a few harder questions:
-
Where does your expertise naturally map within the AI stack?
-
Are you accumulating advantages that are difficult to replicate — technically, geographically, institutionally, or relationally?
-
Is part of your capital allocated to areas you genuinely understand, rather than driven entirely by narratives?
AI will create outsized winners. It will also create sunk costs.
Some people will capture the upside because they already stand near the centre of the system. Others will work hard, but in parts of the economy where the pricing power, learning curve, and capital flow are moving elsewhere.
This is not a moral judgment.
It is a structural one.
The labour market evidence is already pointing in that direction.
The World Economic Forum’s Future of Jobs Report 2025 projected that job disruption could affect 22% of jobs by 2030, with 170 million new roles created and 92 million displaced, resulting in a net gain of 78 million roles. [7]
PwC’s 2025 Global AI Jobs Barometer found that workers with AI skills command an average 56% wage premium compared with workers in the same occupations without those skills. [8]
The IMF has also emphasised that demand for new skills — especially IT and AI skills — is reshaping hiring and wages, but can deepen polarisation if the benefits concentrate among workers and firms already positioned to absorb those skills. [9]
This is why “learn AI” is not a complete answer.
The more precise question is:
Which layer of the AI economy can your learning compound inside?
For some people, the answer may be model workflows or software automation.
For others, it may be semiconductor process integration, supply chain execution, data centre operations, energy infrastructure, or institutional knowledge that becomes more valuable precisely because the system is more complex.
The goal is not to follow every AI trend.
The goal is to find a layer where your judgment improves faster than the noise.
7. Position First. Then Speed.
If there is one way to summarise this, it is this:
The AI industrial revolution is increasingly real.
But for most of us, the important question is not whether we dare to go all-in.
The important question is:
Where do we stand?
Optimism is justified when it is attached to a real position.
Without position, optimism becomes exposure without leverage.
Only after the position is clear does speed become meaningful.
Before that, acceleration may simply mean moving faster in the wrong direction.
Position first. Then speed. Everything else is just motion in the wrong direction.
A Question for Readers
Where are you currently positioned in the AI stack?
And more importantly — why?
Was that by deliberate choice, or did you simply end up there?
About the Author
Writes about semiconductors, silicon photonics, AI infrastructure, and the parts of the technology stack where outcomes are ultimately determined by process windows, but rarely explained clearly in financial narratives.
Research topics span process R&D, advanced packaging, and quantitative research across the AI semiconductor supply chain.
Translating process windows into capital allocation.
Further Reading
-
Next: When Light Replaces Copper — But What About the Process?
-
Upcoming: CPO and Silicon Photonics: The Hardest Layer to Replicate in AI Infrastructure
-
Upcoming: Semiconductors and Packaging: Bottlenecks and Misalignments in the AI Supply Chain
References
[1] Reuters — Big Tech to invest about $650 billion in AI in 2026, Bridgewater says
[2] Federal Reserve Bank of St. Louis — Tracking AI’s Contribution to GDP Growth
https://www.stlouisfed.org/on-the-economy/2026/jan/tracking-ai-contribution-gdp-growth
[3] Federal Reserve — Estimating Aggregate Data Center Investment with Project-level Data
https://www.federalreserve.gov/econres/feds/files/2025109pap.pdf
[4] Taiwan DGBAS — GDP: Preliminary Estimate for 2025Q4, and Outlook for 2026
https://eng.stat.gov.tw/News_Content.aspx?n=2337&s=235721
[5] Taiwan Ministry of Finance — Annual External Trade Report in 2025
[6] TSMC — Q1 2026 Earnings Conference Transcript
[7] World Economic Forum — Future of Jobs Report 2025: 78 Million New Job Opportunities by 2030 but Urgent Upskilling Needed
[8] PwC — 2025 Global AI Jobs Barometer
https://www.pwc.com/gx/en/issues/artificial-intelligence/job-barometer/2025/report.pdf
[9] IMF — Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age
https://www.imf.org/-/media/files/publications/sdn/2026/english/sdnea2026001.pdf
[10] IMF — Global Economic and Financial Implications of Artificial Intelligence
https://www.imf.org/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf
Disclaimer
This article is for industry and technical analysis only and does not constitute investment advice. Any discussion of technology, supply chains, labour markets, or capital allocation is intended as an analytical framework, not as a recommendation of any specific company, security, asset class, or trading strategy.
Hashtags
#AIInfrastructure #AIIndustrialRevolution #Semiconductors #SiliconPhotonics #CPO #AdvancedPackaging #SupplyChain #CapitalAllocation #AIEconomy