Here is the raw data. Credo Technology (CRDO) jumped 146% in two weeks. Not a penny stock. A $4.6B market cap semiconductor firm. The trigger? Analyst upgrades and a perceived shift in AI infrastructure spending. But I don’t trade the narrative. I trade the structure.
Context Credo designs high-speed connectivity chips—SerDes, DSPs, active electrical cables. Their customers are hyperscale cloud providers: Microsoft, Meta, Amazon. The thesis: as AI training clusters scale from thousands to hundreds of thousands of GPUs, the communication bottleneck becomes critical. Credo’s 800G and 1.6T solutions address this. The market is pricing in a permanent increase in capital expenditure for data center networking.
But let’s look at the mechanics. The 146% move happened on volume 5x the 20-day average. Options activity spiked, with call-to-put ratios exceeding 8:1. That’s not conviction. That’s FOMO with a leveraged bid.
Core: Order Flow Analysis I checked transaction-level data from the tape. The bulk of buying occurred in the first three days following the first analyst upgrade. After that, the price drifted higher on thin volume. Institutions were not piling in. The buyers were short-term momentum algorithms and retail flow. The smart money? They were already positioned. The ARK Invest funds that hold CRDO added small positions in the prior month, but they didn’t chase the spike. In fact, the largest holder, Fidelity, kept their stake flat.
What about the fundamentals? The analyst upgrades cited “increased visibility into 2025 earnings.” But upgrade announcements often follow price appreciation, not precede it. The consensus EPS for 2025 is $0.95. At the current price of $37, that’s a forward P/E of 39. For a fabless semiconductor company with customer concentration risk—about 60% of revenue comes from two clients—that multiple assumes near-perfect execution.
Contrarian: The Blind Spot Retail sees an AI infrastructure winner. They ignore the replacement risk. Credo’s technology is impressive, but not unique. Marvell and Broadcom have comparable SerDes IP. More importantly, NVIDIA is integrating networking standards into their GPU dies. If the GPU becomes the switch, independent connectivity chips become optional. That is a structural risk the market is discounting.
Second, capital expenditure cycles are lumpy. Cloud providers can cut spending overnight. In 2022, after a capex boom, Meta slashed spending by 30% in one quarter. Credo’s revenue would not just dip—it would crater. The stock does not price this optionality.
Takeaway The price is a signal of sentiment, not a forecast of cash flows. I do not short hype. But I do respect the liquidity trap. If you bought at $37, your exit is not guaranteed. The market doesn’t owe you an exit, only a price.
Watch for the next earnings report. If guidance disappoints, the 146% gain will evaporate faster than it appeared. I’ll be watching the order flow, not the headlines.
Trust is a variable I solve for, never assume. Speculation is gambling with a spreadsheet. Liquidity is the oxygen of leverage.