
LSCC · Technology
The market has correctly identified Lattice's edge AI positioning and sticky switching costs, but has priced the transformation as complete — what's missing is that the largest piece of the revenue base sits in a geography that is executing, with state backing, on a deliberate strategy to eliminate exactly this kind of foreign semiconductor dependency.
$111.88
$28.00
The switching cost moat is real — gross margins held firm through a revenue collapse that exposed how cyclical the underlying demand actually is. Management quality is in transition, and the incoming CEO's front-loaded compensation package introduces genuine governance uncertainty at exactly the wrong moment.
This is one of the cleaner cash generators in semiconductors — operating cash flow persistently outrunning net income through a brutal down cycle is the clearest possible signal that the earnings quality is genuine, not manufactured. The balance sheet is nearly spotless and the business funds its own growth without diluting shareholders.
The recovery is real and accelerating — sequential momentum, rising FPGA attach rates, and new product families growing fast are not noise. But the structural China substitution risk is happening concurrently with the cyclical recovery, and those two forces are pointed in opposite directions with no obvious resolution date.
Every credible scenario in the DCF produces fair value that is a fraction of the current price — the market is not pricing in a recovery, it is pricing in a transformation at a level of certainty that the data does not support. The multiples are not high relative to history; they are categorically untethered from any reasonable earnings power estimate.
Three concrete threats converge simultaneously: more than half of revenue flows through a geography actively building domestic substitutes under government mandate; hyperscalers have demonstrated willingness to build proprietary silicon for exactly the management plane functions Lattice owns; and the edge AI narrative that justifies the premium faces direct commoditization from dedicated NPU chips. Any one of these is manageable — all three at once, at this valuation, is a serious problem.
Lattice is a legitimately good business wrapped inside a genuinely dangerous valuation. The switching cost moat is validated by the one data point that can't be faked — gross margins that barely flinched through a revenue collapse that would have exposed commodity pricing in seconds. The cash generation is real, the capital-light model is durable, and the companion chip thesis for AI server infrastructure is strategically coherent. The problem is the price, which is not merely expensive but untethered: even an optimistic recovery scenario produces intrinsic value at a steep discount to where the stock trades today, and the market is not pricing in execution risk, it is pricing in certainty. The business trajectory from here is genuinely bifurcated. On one path, FPGA attach rates continue climbing alongside AI server buildout, Avant wins mid-range sockets that expand the addressable market meaningfully, and the edge AI narrative converts from design wins into recognized revenue at scale — that is a legitimate compounding story. On the other path, the China substitution dynamic accelerates, domestic FPGA vendors close the toolchain quality gap under government subsidy, and the business loses its largest revenue geography precisely as it is trying to fund the product investment to win the next cycle. Both of these paths are live simultaneously, and the market is only pricing one of them. The single biggest specific risk is not abstract geopolitical noise — it is the concrete possibility of US export controls expanding to reach low-power FPGAs at the exact price point Lattice owns. A business where more than half of revenue flows through China and Asia is making an enormous, involuntary bet on trade policy remaining stable, and that policy has moved repeatedly and unpredictably in recent years. One regulatory stroke could sever the company's largest growth engine, and no amount of edge AI design wins in the Americas and Europe closes that gap on any reasonable timeline.