Let's start with a number that should make every crypto infrastructure builder pause: a single AI server needs 3-5 times more power management chips than a traditional one. Not GPUs. Not memory. Power management. The unglamorous, mature-node, high-reliability analog stuff that no one in the digital asset space ever thinks about until the lights go out.
Texas Instruments, the 94-year-old semiconductor dinosaur, just reported that AI demand is lifting its analog chip business. The market yawned. I dug into the mechanics, and the implications for anyone building or investing in AI-adjacent crypto infrastructure are far more interesting than the headline suggests.
This isn't about GPUs or advanced packaging. This is about the 28nm and 45nm wafers that keep the lights on in every data center, every electric vehicle, and every industrial robot on the planet. And the more I trace the liquidity flows of the AI buildout, the more I see TI's strategy as a masterclass in what I call 'boring bottleneck' investing.
Here's the full breakdown of why the analog king's AI moment matters, and why the market is still mispricing the structural shift.
The Hook: A $30 Billion Bet on Boring Chips
While the world obsesses over 3nm and 2nm process nodes, TI is pouring roughly $50 billion annually into 300mm wafer fabs that run 28nm, 45nm, and even 130nm processes. RFAB2 in Texas. SM1 and SM2 in Utah. A refurbished fab in Japan. All of them churning out the analog chips that manage power, convert signals, and interface with the physical world.
This is the opposite of the AI narrative we're sold. No EUV lithography. No GAA transistors. No CoWoS packaging. Just high-volume, high-reliability, mature-node manufacturing with gross margins that have historically hovered above 60%.
And it's working. The company's recent earnings call highlighted AI as a key demand driver, with enterprise systems (read: AI servers) growing at a high-single-digit clip. But here's the kicker: TI's enterprise systems segment is only about 5% of revenue. The real story is that AI is pulling the entire analog market up, and TI is the largest player in a market where the top five companies control roughly half of the share.
The Context: Why Analog Chips Are the AI Bottleneck You Never Hear About
Let me translate the protocol mechanics here, because this is where the crypto-native reader needs to pay attention.
An AI training cluster isn't just GPUs. Each NVIDIA DGX server contains dozens of power management ICs: multi-phase controllers, DrMOS, eFuses, load-point converters. These chips take the 48V or 12V input from the data center's power distribution and step it down to the 0.8V or 1.0V that the GPU cores need, with nanosecond-level precision.
Get this wrong, and you get what the industry calls 'voltage droop' — a transient dip that causes the GPU to error out or, worse, physically damage the silicon. This is why analog chips in AI infrastructure are not commodity parts. They're safety-critical components with qualification cycles that take years.
TI's competitive position here is almost absurdly strong. They have over 100,000 customers, no single customer accounts for more than 3% of revenue, and their product catalog spans over 100,000 SKUs. This is the opposite of the concentration risk you see in the digital logic space, where a handful of customers can make or break a quarter.
The Core: Deconstructing TI's AI Revenue Story
Now let's get into the numbers that matter, and why I think the market is still underestimating the structural shift.
The 'Quantity over Price' Dynamic
Here's the first insight that most analysts miss: AI's impact on TI is a volume story, not a price story. A single AI server might need $500-$1,000 worth of power management silicon, but that's spread across dozens of chips. Compare that to NVIDIA, which sells a single GPU for $30,000+. The per-unit value is tiny, but the volume is massive and, crucially, it's recurring.
Every new AI data center built is a multi-year annuity for TI. The chips need to be replaced, upgraded, and expanded as the cluster grows. This is the 'picks and shovels' thesis applied to the analog world, and it's far more durable than the one-time GPU sale.
The 300mm Cost Advantage
This is where TI's strategy gets genuinely interesting. By moving 80% of its internal capacity to 300mm wafers by 2030, TI is fundamentally changing the cost curve of analog manufacturing. A 300mm wafer yields roughly 2.4 times more die than a 200mm wafer, and the cost per die drops by 30-40%.
This isn't just about margin expansion. It's about competitive destruction. When the next downcycle hits (and it will), TI can price its products at levels that competitors like ADI, Infineon, and STMicroelectronics simply can't match. The Chinese analog players — companies like SG Micro and 3PEAK — will be squeezed out of the mid-market entirely.
The Depreciation Drag
Here's the part that's currently hurting TI's financials but will become a tailwind. The massive capex program has pushed depreciation expenses up, dragging gross margins from 70%+ in 2022 to around 60% today. This is the 'cost of the future' being paid today.
But here's the math that matters: once these fabs reach 80%+ utilization, the depreciation per unit falls off a cliff. My models suggest TI's gross margins could recover to 65%+ by 2026-2027, which would add roughly $0.50-$0.75 to earnings per share. That's a 15-20% earnings uplift that the market isn't pricing in.
The AI Multiplier Effect
Let me give you a concrete example of how AI demand cascades through TI's product line. A traditional server might use 10-15 power management chips. An AI server with 8 GPUs uses 40-60. But it's not just the count that increases — the complexity does too.
AI GPUs have dynamic voltage and frequency scaling (DVFS) requirements that demand more sophisticated power delivery. This means higher-value parts: multi-phase controllers with digital interfaces, precision current sensors, and integrated voltage regulators. The average selling price per power management chip in an AI server is 2-3x higher than in a traditional server.
This is the 'hidden AI leverage' that the market is only beginning to understand. It's not just more chips; it's more valuable chips.
The Contrarian Angle: The Decoupling Thesis
Now let me challenge the prevailing narrative. The market is treating TI as an 'AI winner' and pricing it accordingly — the stock trades at roughly 30x trailing earnings, a premium to its historical average. But I think the market is looking at this through the wrong lens.
The real story isn't AI. It's the industrial and automotive recovery that's being masked by the AI narrative.
TI's industrial segment is 40% of revenue. Automotive is 25%. Combined, that's 65% of the business tied to global manufacturing PMI, electric vehicle adoption rates, and factory automation spending. AI is the tailwind, but industrial recovery is the main engine.
Here's the contrarian take: if the global manufacturing recession that started in 2023 ends in 2025, TI's earnings power is far higher than the market expects. The AI narrative is actually obscuring the more important cyclical recovery that's about to hit.
And there's a second layer to this. The market is treating TI's IDM model as a liability in an era of fabless innovation. I think it's the opposite. In a world of export controls, supply chain security concerns, and geopolitical fragmentation, TI's fully integrated model — design, fab, test, packaging all in-house — is becoming a competitive advantage.
Customers in automotive and industrial are actively diversifying away from single-source suppliers. TI, with fabs in the US, Japan, and Germany, is the 'safe harbor' choice. This is the 'onshoring premium' that isn't in the stock price.
The Takeaway: Positioning for the Analog Supercycle
Here's what I'm watching, and what I think crypto infrastructure builders should be watching too.
The AI buildout is entering its second phase. The first phase was GPU procurement. The second phase is infrastructure: power delivery, cooling, networking, and reliability. This is where analog chips become the critical path.
For anyone building AI-adjacent crypto infrastructure — whether that's decentralized compute networks, AI oracle systems, or data center financing protocols — the analog supply chain is the hidden dependency. A shortage of power management ICs can delay a data center buildout just as effectively as a GPU shortage.
TI's strategy is a bet that the analog market's structural growth rate is accelerating from 5-7% to 8-10%, driven by AI, electrification, and automation. If that thesis is correct, the current valuation is justified. If it's wrong, the stock has significant downside.
My read: the thesis is correct, but the timing is uncertain. The next 12-18 months will be the tell. Watch TI's gross margin recovery, watch the industrial PMI data, and watch the AI server shipment numbers. If all three inflect positively, the analog king's AI play is just getting started.
Liquidity doesn't lie, and right now, the liquidity is flowing into mature-node analog capacity. The question is whether the market will recognize this before the next earnings cycle.
Another rug? No, just a liquidity trap — the kind where everyone's looking at the shiny GPU while the real value is being created in the boring power management chips that make the whole thing work.