The Zhitong Finance App learned that Morgan Stanley released a research report saying that after having a one-on-one conversation with Chief Operating Officer Jeff Clarke at Dell headquarters, the bank had a deeper understanding of Dell's bullish logic, but it still maintained a “neutral” rating and a target price of $511. The central debate in the report is whether AI agents are bringing about another round of hardware cycle peaks or years of structural prosperity in local infrastructure. Morgan Stanley made it clear that the meeting made it re-evaluate its previous partial judgment, and believes that the probability that the $756 bull market scenario will be realized in the next 12 months has increased.
This time is different: from refresh cycles to TAM expansion
Clarke emphasized to Morgan Stanley that the hardware market has experienced 8 boom and bust cycles over the past 40 years, but these cycles are mainly driven by fluctuations in refresh rate due to changes in the per capita penetration rate of equipment, rather than structural expansion of potential market size (TAM). For example, during the pandemic, the PC refresh cycle accelerated due to an increase in the ratio of PCs to people, but the ratio normalized after the social restart, and PC TAM was not greater than before the pandemic.
The difference of AI intelligence is that for the first time, it decouples cognitive output from the number of employees, so that companies can complete more work without increasing manpower, and productivity can be increased by 10 times or even 100 times. Dell believes this return is clearly visible, and businesses of all types will follow this path. By 2030, inference tokens are expected to grow 87 times, add 200 gigawatts of computing power, and grow ZettaFlops to 830. Since smart devices require 1 billion more tokens to work than basic chatbots, token growth will force data centers to restructure faster computing and higher CPU:GPU ratios, and catalyze a structural rise in computing and storage requirements.
Dell also believes that AI agents will be a hybrid deployment model. Businesses dynamically distribute inference load between public clouds and local facilities based on security and cost trade-offs. Dell itself will place content-related workloads in the public cloud, but will never move proprietary source code or telemetry data out of Dell facilities. Considering that inference tokens will grow 87 times by 2030, “even taking only a small chunk of a large pool will form a very large number.” The popularity of open source weighting models will also drive enterprises to increase investment in local infrastructure to optimize costs.
Servers and storage: increased density coexist with supply bottlenecks
Shipments of traditional servers are still declining year over year, but Dell believes this is not about weak demand, but rather that servers are becoming more powerful, more dense, and more efficient. 17th and 18th generation Dell servers can replace up to 13 traditional 14th generation servers. In the short term, the number of server units has declined but ASP has risen; as data centers are rebuilt with accelerated computing as the core and the popularity of intelligent devices boosts demand for CPU servers, more intensive server units will eventually resume growth. Clarke even pointed out that if the inference token actually grows 87 times within 5 years, the growth of traditional server units may also become exponential. Storage also benefits. Actions such as KV Cache and intelligent object retention will bring greater data storage and knowledge retention requirements.
Supply is Dell's biggest challenge. Clarke said that the historical commodity cycle usually lasts 2 to 4 quarters, driven by mature hardware replacement cycles or boom/bust decisions; however, if you believe the growth algorithms of gigawatts and tokens, memory, HDD, etc. are not a short-term cycle, and the shortage may continue for another five years. When asked what keeps him up at night, Clarke's answer wasn't macroscopes, geopolitical conflicts, overbuilt data centers, or power shortages, but supply. He used “We're in Neverland” to describe the current environment, and Dell is assuming there will be a long-term shortage of critical components.
Profit margin controversy: not a simple “profit superposition”
Morgan Stanley previously pointed out in an industry report that the infrastructure gross margin of Dell and HPE (HPE.US) has experienced an unprecedented expansion, which is mainly due to “profit superposition,” that is, using rising memory costs to obtain higher profits under rigid demand. Clarke acknowledged the year-over-year expansion in server and storage margins, but provided a different perspective: in a world where bit supply is limited, companies will allocate limited parts to the highest gross margin products; at the same time, the structural growth of infrastructure TAM has reduced competitive pricing intentions for incremental customers. As TAM grows, instead of selling servers with narrow profit margins to acquire new customers, it is better to maintain stricter pricing discipline and raise the price bottom. Clarke believes the increase in server and storage margins brought about by this is structural.
In terms of operating expenses, Dell's internal AI modernization is in full swing. Clarke emphasized that Dell is using AI to improve productivity and efficiency in supply chain optimization, customer service, coding, sales, and R&D. These investments won't slow down anytime soon, as Dell is reaping “huge benefits” from them, so operating expenses are unlikely to drop year over year. However, as revenue grows faster than investment, Dell will still release operating cost leverage.
Valuation differences: long and short sides face off for a long time
Morgan Stanley pointed out that investor exchanges showed that both long and short agreed that Dell's revenue and profit growth over the next 18 months will be extremely strong. Buyers' expectations for FY2028 earnings per share have risen to the $40 to $50 range. Morgan Stanley's forecast is $36.47, and the market consensus is $30.44.
The real difference is the persistence of growth: many parties agree with Clarke's worldview and are willing to give FY2028 earnings per share 15 times the price-earnings ratio, corresponding to the share price of 600 to 750 US dollars; the empty side believes that this is another boom/bust cycle, and the AI value is more in the cloud. When demand rigidity shifts to demand collapse, Dell's profit may be halved, so they are only willing to give the 2028 peak earnings per share a price-earnings ratio of 8 to 10 times the price-earnings ratio, corresponding to the share price of 320 to 500 US dollars.
Morgan Stanley admits that previously it was closer to the view of a bear market, but Clarke's meeting prompted the bank to re-evaluate, focusing not on token growth itself, but on whether this growth will actually lead to long-term prosperity in local infrastructure. Currently, the bank still maintains a “neutral” rating and a target price of $511, but believes that the $756 bull market scenario is more likely to be realized in the next 12 months. The bull market scenario was $756, based on 18 times the 2028 bull market earnings of $42 per share; the benchmark scenario was $511, based on 14 times the 2028 earnings per share of $36.47; and the bear scenario was $306, based on 12 times the 2028 bear market earnings of $25.53 per share.
In terms of financial models, Morgan Stanley expects Dell's revenue to grow 93.9% in FY2027, 33.1% in FY2028, and fall 4.7% in FY2029; non-GAAP gross margin will fall from 20.4% in FY2026 to 16.0% in FY2028 and 16.2% in FY2029; operating margin will rise from 8.8% to 11.1% in FY2027, then fall back to 10.0% and 10.2%; free cash flow will rise from US$8.552 billion to $21.174 billion, and $28.876 billion in 2029 $279.10 billion. The bank's revenue forecast for the 2028 fiscal year was US$293 billion, higher than the market consensus of US$23.4 billion; the earnings per share forecast was US$36.47, higher than the agreed US$30.02; and the net profit forecast was US$22.813 billion, higher than the agreed US$19.269 billion.