Want to hear an uncomfortable truth that Wall Street doesn't want to admit? The AI supply chain crisis can't be solved by money. | This week let's examine the immovable constraints that collide with AI's unstoppable growth, which are bound to influence the next decade of returns. |
|
| | The Ceiling Nobody Saw Coming | Semiconductor foundries can't work overtime because the basic unit (wafers per month) is fixed at a brutal level. It takes 12 to 16 weeks to process a wafer at advanced nodes, such as 3nm and 5nm. Combine this with the AI chips occupying more wafer space, and we get limited chips per wafer. | Here's some info that keeps AI investors awake at night: TSMC has publicly acknowledged its advanced-node capacity is three times (approximately) smaller than current AI demand. | Additionally, moving to a new node, such as 2nm, reduces early yields to as low as 10-20%. Until yields reach 70-80%, volume output will be limited and very expensive. | Despite billions of dollars in capital spending, the AI chip supply is constrained, marking capacity as a physical ceiling rather than a budget item. |
|
| | | | | Monday Algo Spots Major Stock Breakout… | A 89.5% gain on Nuburu Inc… | A 149% gain on MSS… | A whopping 536% gain on Rebel Holdings… | These are stock winners you could have already locked in the last few months, simply follow the signals in my new algorithm: The Monday Algo. | You see, every Monday like clockwork, my algorithm spots stocks that are about to break out for gains like +149%, +190%, and +536%... in as little as one day. | Now, it's flashing buy on a new dirt-cheap stock with a great chance of moving +100% or more this Monday. | And if I'm right, this stock will move FAST… | So get the details ASAP…
And join me and my students every Monday for a shot at a full year of profit in just one day. |
| |
| | |
| The Timeline That Won't Budge | A leading-edge fab can be built with $20 billion in capital expenditures and $2 billion in annual operating costs. The expense is bearable, but time? | | Add another 6 to 12 months for tool installations, and a fab (announced today) can ship meaningful volume in 2028 or after. |
|
| | The Power Problem That's Getting Worse | A single advanced fab requires 1.5 to 2.5 billion kilowatt-hours per year – enough to power 200,000 homes. | The reason for the hunger? EUV lithography tools. Each one consumes several megawatts. They also require constant cooling and power stability. | However, the most concerning aspect of all this is the wait time for grid connection. In the U.S., it stretches from 5 to 7 years. The country's grid spare capacity is expected to drop below 15% (from 26%) by 2030, which could affect fab run speeds, potentially even after construction is complete. |
|
| | The Water Crisis Nobody's Talking About | Like electricity, semiconductor fabs require massive amounts of water. | A leading-edge fab uses approximately 5,000 liters of water per wafer, with annual consumption hitting 3 to 4 billion gallons. | In Hsinchu, TSMC's fabs alone used more than 150,000 tons per day. That's about 10% of the city's entire supply. | Do you remember the drought in Taiwan in 2021? Trucks carried water to the fabs to keep them running. The operational reality is that by 2040, 40% of fabs worldwide will be located in high-water-stress regions. |
|
| | The Material Squeeze Tightening Quietly | Material constraints also have a say in the AI supply chain. | Copper is expected to have a deficit of nearly a ton in refined production in 2025, primarily due to data centers. With China as its primary supplier, Gallium's 2,100-ton-per-year supply volume carries geopolitical risk. Meanwhile, the limited capacity of ABF substrates is anticipated to expand further. | By 2035, the deficit in copper supply is expected to affect approximately 32% of global semiconductor production. | Material shortages raise costs and slow down tool delivery, thus affecting the AI supply chain. |
|
| | The Geographic Concentration That Terrifies Me | AI chip production is dangerously concentrated with: | Approximately 90% of advanced chips made in Taiwan. 40% of all fabs will be located in high-water-stress regions by 2040. TSMC is projected to account for 12% of Taiwan's economy by 2030.
| Some significant regions are facing equal risks. Arizona, for instance, faces a long-term drought that threatens its water supply. | You cannot find a spare when a region is hit with disruption, because there isn't any. Hence, a single major event could halt global AI hardware production. |
|
| | Why Government Money Won't Save Us | The U.S. CHIPS Act and similar programs offer monetary support for fabs. | But there is a problem. Money can't fix the shortage. | The U.S. is expected to face shortages of more than 300,000 engineers and 90,000 skilled technicians over the next decade. Environmental permits can take 12 to 18 months to obtain before construction even begins. | Government support helps, but it can't fix the multi-year lag in announced projects. |
|
| | Your Investment Cheat Sheet | Electricity, water, materials, and time constraints limit AI growth as effectively as ideas and software. | Structural Winners: | Foundries with operating advanced capacity (TSMC primarily), since no alternatives exist. Equipment makers supplying EUV, inspection, and deposition tools. They have years of backlog. Infrastructure companies related to power, water recycling, and grid upgrades.
| Structural Losers: | Industries pushed out of capacity, including automotive and industrial electronics. Smaller AI companies are unable to secure packaging or HBM allocation. Regions lacking infrastructure, like power or water.
| Key Signals to Watch: | Rising wafer price – It indicates foundries retain pricing power. Yields above 70% at new nodes – Suggests cost per chip falls & volume increases. Grid connection approvals – It means reduced expansion risk. Water recycling above 70% – Points to sustainability in operations.
|
|
| | The Bottom Line | Companies can control AI growth and its pricing by controlling physical capacity. | Investors must understand the limits, such as power grids, water pipes, and the laws of physics, to identify long-term winners and avoid feeble growth stories. Physical constraints exist, and if they continue to hamper supply for years, the likely winners of the decade will be those who own operating capacity today. | Money can't buy you time, and time is the one resource no one ever gets enough of. |
|
| | | | | Important disclosures: This newsletter is provided for informational purposes only and does not constitute investment advice. All investments involve risk, including possible loss of principal. Please consult with your financial advisor before making investment decisions. |
| |
| | |
|
|
No comments:
Post a Comment