Credo Technology Group Holding Ltd (CRDO) stock intrinsic value fair value estimate & owner earnings analysis

Credo Technology Group designs high-speed connectivity chips and cables used to link servers inside data centers, especially the specialized clusters that train and run artificial intelligence models. Its core products are Active Electrical Cables (AECs) and optical/electrical signal-processing chips (SerDes-based DSPs) that let data move between racks of AI hardware with lower power and higher reliability than older cable technology. Credo both licenses its underlying chip-design technology and sells finished hardware, and in fiscal 2026 it acquired two smaller companies -- Hyperlume, a maker of miniature LED-based optical links, and, just after year-end, DustPhotonics, a $770 million all-cash silicon-photonics deal -- to extend into next-generation optical interconnects. Credo is legally incorporated in the Cayman Islands, a common holding-company structure for many US-listed companies, but is managed and run out of San Jose, California, with manufacturing outsourced to contract manufacturers in Asia -- a standard fabless-semiconductor supply chain, not an indication of Chinese ownership or control. The business has been transformed over the past two years by the buildout of AI computing infrastructure. Revenue grew from $193 million in fiscal 2024 to $437 million in fiscal 2025 to $1.34 billion in fiscal 2026, and the most recent quarter continued to grow strongly, though at a slower pace than the prior year. Cash profitability -- after treating stock-based compensation as a real cost and after the reinvestment needed to support growth -- was break-even to negative from fiscal 2020 through fiscal 2024, then turned solidly positive in fiscal 2025 and fiscal 2026. We anchor the valuation on the average of those two most recent, transformed years rather than on the smaller pre-2025 business, which was a materially different, sub-scale company, or on the single highest year alone. A defining fact about Credo today is customer concentration: one customer accounted for 53% of fiscal 2026 revenue, and the top ten customers made up about 90% of revenue. Design wins with these customers are not binding long-term contracts -- the company states plainly that a win can take a year or more to turn into revenue, and if new wins are not secured, revenue in later years can decline. That combination -- extraordinary recent growth built substantially on a small number of customer relationships that are not contractually locked in -- is the central swing factor here, and it is reflected in materially different assumptions across the three cases below rather than a single forecast. Base case (roughly $84/share): growth decelerates sharply from the pace of the last two years -- a large step up in the coming year, still supported by shipments already occurring and the unfulfilled backlog cited in the filings, followed by progressively slower growth as the current customer relationships mature, settling into high-single-digit growth by the back half of a ten-year window. We do not extend the recent design-win record indefinitely, since the company itself discloses those wins are not contractually secured; ongoing R&D investment (which nearly doubled in fiscal 2026 to $279 million) is treated as sustaining, not growing, the earnings level once the current ramp runs its course. Bear case (roughly $38/share): the concentration risk above materializes -- the largest customer pauses or reduces orders as AI infrastructure spending, which the industry has repeatedly shown to be lumpy, cycles from build-out into digestion; the new silicon-photonics acquisition fails to gain design wins and its cost is not recovered; growth stalls and modestly reverses for several years before stabilizing at a lower, more diversified base. Bull case (roughly $205/share): current AI-infrastructure demand continues at a strong pace for longer, the newly acquired silicon-photonics technology succeeds in winning a place in next-generation optical interconnects, extending the addressable market beyond today's electrical cables, and the customer base broadens as the company itself expects. The company carries no debt and, even after paying $770 million cash for the DustPhotonics acquisition in the most recent quarter, still holds roughly $764 million of cash and short-term investments -- a straightforward net-cash balance sheet that adds a little over $4 per share in every case above and needs no adjustment for financing risk. Because more than half of revenue rests on one customer relationship that is not a binding contract, and because $860 million was just committed to two acquisitions in unproven technology, the margin-of-safety price is set at a large discount to fair value. Our margin-of-safety price is around $38 per share -- close to the bear-case value, so that even if the more difficult scenario plays out, the price is not far from what the business would still be worth. One item worth flagging: a portion of prior-year revenue was reduced by a non-cash customer-warrant charge (tied to warrants issued to a large customer) that fell to zero in fiscal 2026, which makes the reported revenue growth rate look somewhat better than underlying volume growth alone would suggest. We addressed this by basing the valuation on cash-flow-derived earnings, which are not affected by this accounting item, rather than on the reported revenue growth rate.