Piper Sandler analysts David O'Connor and Zackary Altman initiated coverage on five artificial intelligence semiconductor companies on Friday, assigning overweight ratings and price targets that imply between 19 percent and 37 percent upside from the September 10 close. The call rests on a simple supply argument: AI chips are largely sold out for 2025 and order books are already full for 2028, while hourly pricing for Nvidia's four-year-old H100 GPUs has climbed 30 percent from $2.00 in December to $2.60 currently.
The compute market math
The analysts project the broader compute market will expand fivefold to $2.2 trillion by 2030, with roughly $2 trillion flowing to AI chips and another $200 billion to central processing units. That top-down figure frames the individual price targets. Nvidia carries a $300 target, about 37 percent above the September 10 level, supported by an estimated 80 percent market share and a forward price-to-earnings multiple of 14 times fiscal 2028 estimates. Broadcom's $460 target implies 27 percent upside; the firm cites a 75 percent share of the inference ASIC market where demand is running double supply, and a 19 times one-year forward multiple that Piper calls the cheapest in the group.
Valuation dispersion
Arm Holdings received a $320 target, 26 percent above its current price, though the stock trades at a forward P/E of 114 times as the company begins manufacturing its own server CPUs while still licensing IP to customers including Nvidia, Alphabet, and Amazon. Advanced Micro Devices and Marvell Technology both carry 19 percent upside, AMD at a $600 target tied to agentic AI and inference GPU deals, Marvell at $270 on what Piper describes as a transformational deal with Alphabet. The source text ends mid-sentence on Marvell's catalyst.
What the market is pricing
The spread between the highest and lowest implied returns, 18 percentage points, reflects divergent views on moat durability. Nvidia's CUDA ecosystem and Broadcom's custom silicon franchise command confidence; Arm's transition to a chipmaker and AMD's share gains against Intel are treated as more speculative. The common thread is a supply constraint that has turned four-year-old hardware into an appreciating asset, a dynamic that rarely persists without inviting new capacity or demand destruction.
