Johan Ekroth
AI Infrastructure/Space/Defense Investor, Bottleneck Research
Arm is probably the cleanest structural play on the CPU demand surge from inference and agentic AI. Intel CEO Lip-Bu Tan says the CPU-to-GPU ratio for these workloads is now 4:1. @ren_stocks notes Nvidia's Jensen Huang said its Vera CPU opens a $200B new market, while agents need CPUs for orchestration between inference calls; server CPU spending is projected at $170B-$211B by 2030. x.comÂ
Arm Holdings looks unattractive because its nearly 300 P/E ratio is difficult to justify against only 22% year-over-year revenue growth. x.comÂ
Arm's forward P/E near 190 already prices in much of its expected growth, leaving a limited margin of safety. Its fundamentals are sound and AI and data center demand is strong, but a long-term position is better taken after a valuation pullback or further earnings delivery. x.comÂ
Arm remains a long-term winner over the next 10 years, but not a short-term buy after running too far, too fast. Its performance has to catch up with at least part of the move, so buying at these levels carries elevated risk if a negative surprise hits. x.comÂ
Arm's valuation multiple is too high to be appealing. x.comÂ
Arm remains wildly overvalued even after falling 175 points from its top, still trading at 336x earnings, 236x forward earnings and 61.5x sales. Eventually the bids will stop, there will be no bailout rally and chasers will be forced to learn the fundamentals. The price target remains $100-120, reached slowly and then all of a sudden. x.comÂ
SoftBank owns about 90% of Arm, leaving a locked-up float that looks almost certain to be pushed far higher in the early innings of a GPU mania. The bet is not on Arm's technology but on others squeezing that tight float. x.com x.comÂ
Arm is well positioned for the coming CPU boom, but its stock already prices in the next five to six years. It dominates AI ASIC servers, and ASICs' share of servers should grow substantially as AI workloads shift from training to inference. Yet at about 30x 2031 earnings it is optimistically valued and would need a lot of upside surprise to become attractive. x.com x.com x.comÂ
Arm's post-earnings rise makes little sense at over 40x sales, because the results did not justify it. Remaining performance obligations fell and consensus numbers saw only very modest adjustments, yet the shares rose after successive double-digit moves. x.comÂ
Arm's fundamentals remain extremely strong, but component shortages could cost it margin as it moves into manufacturing. It confirmed accelerating demand of up to $2B for its new AGI CPU, yet has components for only about 50% of it, so meeting demand may mean paying higher prices. Management expects CPU demand to rise about 4x by 2030, which means about 4x the component needs too. At a valuation priced for perfection, the market may not accept any hit to growth or margins. x.com x.comÂ
Arm cannot even raise its $2 billion customer-interest figure in what should be the greatest CPU market of our lifetimes, and whether it can deliver even that is still unproven. x.comÂ