Copper isn’t dead in the AI arms race. And a veteran Goldman Sachs analyst believes one of the biggest winners in the race faces more growth ahead.
Credo Technology (CRDO) is not an Nvidia (NVDA) — at least not yet. The stock often gets overshadowed by larger players in the AI race. However, even after surging 180% in 2025 on exploding data-center demand, Credo just got a major Wall Street endorsement.
Goldman started coverage with a buy rating and a $165 price target, which represents approximately 27% upside from recent levels.
In the note, Goldman made its thesis clear.
The call is significant because it goes to the heart of the issue for one of AI infrastructure’s biggest debates. The question is, will optical networking quickly replace copper, or does copper still have a multi-year runway?
Goldman has made its decision. It’s in the second camp.
Goldman just made a bold call on a surging AI stock.Photo by Piaras Ó Mídheach on Getty Images ·Photo by Piaras Ó Mídheach on Getty Images
Credo’s proposition is unique in an overcrowded AI market. Compared to Palantir, a software company, Credo’s approach is a strategic advantage, since it is a pick-and-shovel play.
Credo’s business is bedrocked on the cables that link AI servers within hyperscale data centers.
Credo has the most market share in high-speed Active Electrical Cables.
These cables connect GPUs inside AI clusters and plug into racks of servers, the very infrastructure that makes Nvidia’s chips — and the chips of large companies including Amazon, Microsoft, and Meta — work.
At that scale, connection speed and reliability are becoming critical to the mission.
Goldman argues that for short-range connections inside racks and between adjacent racks, copper-based AECs offer an ideal mix.
They are less expensive than optical options.
Copper-based AECs can reduce power use by up to 50%.
They offer high signal integrity.
Copper-based AECs minimize “link flaps,” which are short breaks in the connection that can stop an AI cluster from training.
Even a short break can be very expensive when working with huge AI workloads that can run for days.
A longstanding debate dominating the stock markets is how long copper can compete against optical solutions. Goldman says the transition is slower than initially feared.
About 80% of data-center switching ports are expected to remain at speeds where copper solutions are still useful until 2030, based on industry forecasts, Goldman added. It forecasts that the move to higher lane speeds will be slow, even as speeds rise, keeping copper useful “until at least 2032.”
More AI Stocks:
That’s an important fact to note, considering that more than 90% of Credo’s revenue currently comes from its HiWire AEC products.
Goldman also acknowledged the competition; Marvell and Astera Labs are gaining on Credo, but it has an ace up its sleeve. Credo’s vertically integrated model and manufacturing strategy support a favorable price-performance edge versus its peers.
Goldman is also pointing toward strong customer traction. Credo is producing revenue at four of the top five U.S. hyperscalers, and a fifth is beginning to ramp.
The fundamental momentum is hard to ignore.
In its most recent quarter, Credo posted revenue of up to 272% year over year to $268 million, with an EPS of 67 cents racing ahead of expectations and leading to shares increasing 10% on the report.
The $165 price target is based on a 26x forward earnings multiple. That is well below Credo’s historical median multiple of 61x and near the area that Goldman calls a trough valuation.
In short, the firm sees earnings power rising while valuation remains under pressure.
Goldman isn’t ignoring the risks that could impact Credo.
The biggest threats include:
Faster-than-expected optical adoption
Intensifying AEC competition
Customer concentration (more than 90% of revenue comes from four hyperscalers)
If hyperscaler capex slows, the stock will come under pressure. Another scenario that could impact Credo is the aggressive shift toward optical networking.
Still, Goldman’s bull/base/bear DCF framework suggests a positive skew of 1.6:1, reinforcing its favorable risk/reward stance.
ZeroFLAP optical transceivers aimed at preventing link failures
Active LED Cables, which use microLEDs and could double the addressable market versus AECs
Memory gearboxes made to help AI inference bottlenecks
Management believes that its total addressable market could grow to $10 billion in the next few years, which is more than three times what it was 18 months ago.
But for now, the main story stays the same.
AI buildouts are galloping along. Goldman Sachs forecasts that hyperscaler capex could reach more than $533 billion in 2026. And inside those huge GPU clusters, speed and reliability of connections are everything.
If copper really does stay “stronger for longer,” Credo may still have more to climb.
