The market is increasingly seeing a peak in computer chip pricing and treating earnings of manufacturers accordingly.The forward multiples of memory chip names in particular are low, with Micron now around 7x forward earnings but the market is sensing that the peak is close and that memory prices will do what they always do: Revert to the low-margin mean. Shares of the company are down 27% from their post-earnings peak last month.In that earnings report, the company forecast earnings of $31 in the current quarter. That puts that price at just 29 quarters of earnings or 7.3x years and still accelerating. That's dirt cheap if you believe the company can continue to extract extreme margins for 2-3 years but the market is questioning that. Commodity-stock watchers know that companies are actually the most-expensive is when multiples are cheapest.One of the reasons shares peaked was talk like this, which was from AI-watcher Andrew Curran:I'm posting this prediction now so I can quote it later. There has been a significant breakthrough in architecture - specifically around memory efficiency - not by one of the big labs, but by a team that was spun out of OpenAI (not SSI). They will probably announce it soon.We are still waiting on that but it appears the rumors have legs.Aside from the 8.2% drop in Micron today, here are now the memory names look:WDC -9.6%SNDK -12.4%Other chip names are also getting beaten up:INTC -10.5%AMD -8.1%GLW -7.7%MCHP -6.1%TXN -5.3%NVDA -2.0%What's incredible is that despite that rout and a 0.6% decline in the S&P 500, it's concentrated. There are 151 stocks down and 326 higher. In the Nasdaq 100, it's 30 down and 49 up.So the market is filtering through the AI noise in creative ways and separating out some of the overall impact of the technology and on the economy. It seems as though there is no doubting the overall impact of AI but the last month has been an exercise of who is going to be reaping the profits. Ultimately, I believe that has to trickle down to real world companies and real world use cases or none of it makes any sense.Here is a note from Morgan Stanley on the current state of the memory trade:Memory – Changing TidesWe address our most debated memory topics with investors: (1) When the biggest spender of AI reportedly has compute to sell, does the story begin to change? (2) Why are stocks not re-rating post LTAs? (3) Peak rate of change vs an extended memory cycle.Peaking rate of change ... Memory is still a cyclical industry – although different this time – and is approaching peak rate of change in (1) pricing YoY, (2) inventory, (3) the earnings revision breadth. This can reach a point where the market begins to go sideways and is also a reason why memory can take a break here. We have had three cyclical corrections in the memory space since Generative AI (Nov 2022), and would view corrections in the context of a structural bull market for AI capex.... but not the end of the cycle. The market moves on less good or less bad news. Selling excess AI compute is the new debate, with “chipflation” potentially pushing hyperscalers out of the frontier AI race versus our bullish view on monetizing the AI stack and AI demand/token penetration that is still in place.Volatility in the most crowded area of the market. Some investors are unwinding positions as recent volatility in memory is making historically high exposure levels in the space harder to maintain. Our investor conversations over the past week reflect sensitivity to this dynamic and growing interest in broadening laggard opportunities. Memory momentum is likely to fade further with EPS revisions breadth for the group against historical highs and the hyperscalar stocks under pressure.What we are watching – corporate earnings. How the market reacts to earnings and what hyperscalers have to say matters more for share prices at this stage than comments from memory companies, which are likely to stay bullish at this point in the cycle. For the AI spenders, token maxing should help 2Q earnings, but less robust 3Q guidance relative to market expectations is another debate, given token minimization, competitive low cost open source LLM, and chipflation impacts on margins.Bottom line. We remain bullish longer term based on earnings growing over 35-40% in 2027e and the ramp of Agentic AI (Rise of the AI Agent – Global Implications). But in the near term, we are increasingly expecting some share price weakness ahead of earnings reporting. Our stock preferences are where the money is being spent and bottlenecks emerging, favoring DRAM and legacy memory over NAND, with least preferred are memory module makers.So it's back to waiting for earnings from the hyperscalers. This article was written by Adam Button at investinglive.com.