Wall Street's analyst desks moved decisively this week in response to company results, IPO market activity and changing assumptions about memory pricing, each development carrying implications for the AI investment narrative.
Micron: Susquehanna re-prices a memory leader off exceptional results
Susquehanna increased its price target on Micron Technology to $2,000 from $1,750 following a fiscal third-quarter report that outpaced consensus across sales, profitability and forward guidance. Micron posted revenue of $41.46 billion and adjusted earnings per share of $25.11, both figures comfortably above Street expectations of $35.91 billion and $20.86 respectively. Gross margin came in at 84.9%, exceeding prior consensus of 80.8%.
For the current quarter Micron provided guidance around $50 billion in revenue and EPS near $31, again ahead of consensus forecasts. Susquehanna cited as a structural inflection point Micron's 16 signed strategic customer agreements. These non-cancellable, take-or-pay contracts typically span five years and currently cover roughly one-fifth of DRAM bits and one-third of NAND bits. Micron projects that, when fully executed, more than half of its revenues will be governed by such agreements.
The contracts contain price floors, which Susquehanna interprets as protections that could secure gross margins above prior-cycle peaks even under adverse demand scenarios. Building on those terms and the stronger-than-expected operating performance, Susquehanna now expects Micron to generate free cash flow in excess of $110 billion in fiscal 2027, with the firm projecting annualized EPS potentially reaching $200 as Micron exits FY27.
Operational detail within the quarter supported the upgrade. Micron's DRAM revenue more than tripled year-on-year to $31.3 billion, with HBM4 ramping at twice the pace of the prior generation and customer commitments extending into 2028 and beyond. NAND revenue nearly quadrupled to $9.9 billion, and data center solid-state drive revenue topped $5 billion in the quarter alone. Micron also raised its outlook for the high-bandwidth memory total addressable market, now expecting HBM to surpass $100 billion in 2027, a year earlier than previously projected.
SpaceX: Argus opens coverage with Hold, flags valuation risk
Brokerage house Argus initiated coverage of newly public SpaceX with a Hold rating, arguing the IPO's record setting valuation will require time to be justified by the company's operating results. SpaceX completed an initial public offering earlier this month that raised total proceeds of $85.7 billion at a valuation above $1.75 trillion, marking the largest IPO in history. That valuation implied a price-to-sales multiple of roughly 95 times projected 2025 revenues.
Argus analyst Steve Silver underscored concerns around SpaceX's valuation and its path to consistent profitability. The firm described SpaceX's operating plan as combining traits of a mature infrastructure business and a venture-style growth investment, a blend that has not yet produced steady profitability. Since the listing, shares exhibited sharp intraday swings, rallying as much as 67% above the offering price before retracing to trade around 10% above the IPO price. Argus indicated ongoing volatility should be expected given a tight share supply, the company's early inclusion in major equity indices and the prospect of upcoming lockup expirations.
While the initial stance is cautious, Argus left open the possibility of upgrading the coverage should the stock fall for nonfundamental reasons or if revenues and earnings accelerate materially faster than currently anticipated.
Qualcomm: Morgan Stanley lifts stance after ambitious data center guide
Morgan Stanley moved Qualcomm's rating from Underweight to Equal-weight and raised its price target to $231 from $146 after the chipmaker used its investor day to outline an aspirational data center revenue target of $5 billion for fiscal 2027. Analyst Joseph Moore acknowledged the bank's prior skepticism and noted the $5 billion target is at least twice the amount Morgan Stanley had expected.
Moore was more guarded on Qualcomm's longer-term aim of $15 billion in data center sales by fiscal 2029, characterizing that goal as more aspirational and highlighting execution risks around accelerators and server CPU introductions. He noted that Qualcomm's planned mid-2028 entry into the CPU market could face a different competitive reception today than it would have earlier, given rapid capacity expansion from established suppliers including NVIDIA, AMD, Intel and cloud providers developing custom silicon.
The analyst stopped short of a full Overweight recommendation, citing superior value in dominant, established players and persistent headwinds in the smartphone market. Nonetheless, Morgan Stanley found the upgrade justifiable, pointing to Qualcomm's credible management and the company's relative underperformance this year when compared with other AI beneficiaries.
UBS: Taking profits in semiconductors, rotating to defensive AI exposures
UBS reduced its combined semiconductor and hardware weighting in its AI strategy from roughly 76% to around 61%, crystallizing gains after a Micron-driven rally that pushed the SOXX exchange-traded index up 4% in a single session and about 10% across June. The bank said profit taking focused on small- and mid-cap companies within the AI supply chain, spanning optics, baseboard management controllers, substrates, cooling systems, advanced packaging and analog components.
Despite the trimming, UBS noted its AI strategy still maintains an overweight position versus the Nasdaq 100, equivalent to roughly 20-25 percentage points relative to the index's roughly 42% exposure to semiconductors and hardware. Concurrently the bank increased exposure to more defensive sectors of the AI ecosystem, moving that allocation to about 20% from less than 1% earlier in June. This defensive sleeve targets data center operators, telecommunications companies and select payments firms with prudent balance sheets and stable dividend profiles.
UBS emphasized that the medium-term AI demand case remains intact, supported by cloud acceleration and agentic AI workloads, but warned of a scenario in which a material decline in the share prices of hyperscalers could pressure their management teams to scale back capital expenditures. Within semiconductors, the bank continues to prefer foundry exposure such as TSMC, semiconductor capital equipment names including Applied Materials, and memory suppliers like SK Hynix. UBS also said it maintains a significant underweight to the so-called Magnificent Seven, with those names representing about 18% of its portfolio weight.
Bernstein: Memory targets raised as HBM price moves threaten to raise AI capex
Bernstein updated price targets upwards for three major memory suppliers on the view that HBM pricing needs to rise materially to close a profitability gap with conventional DRAM. The broker raised its target on Samsung to KRW440,000 from KRW225,000, on SK Hynix to KRW3,300,000 from KRW1,150,000, and on Micron to $1,300 from $510, maintaining Outperform ratings on all three. Bernstein kept an Underperform rating on KIOXIA, which it notes has no HBM exposure.
Bernstein's central contention is that since the third quarter of 2025 conventional DRAM prices have increased roughly 4.5 times while HBM prices, typically locked by annual contracts, have remained largely unchanged. That divergence has meant that deploying capacity to conventional DRAM currently yields more than twice the revenue and nearly three times the gross profit per wafer compared with HBM.
To narrow this gap, Bernstein models a two to 2.5 times increase in HBM prices next year. The broker reasons that supply-side participants and GPU/XPU companies are actively negotiating 2027 HBM pricing because keeping HBM affordable is strategically important for the broader AI ecosystem.
Bernstein warned that rising HBM prices could feed through the supply chain and raise AI infrastructure costs. Unlike conventional DRAM and NAND, HBM is incorporated directly into GPUs; hence pricing moves for HBM are embedded in the cost base of chip suppliers. Bernstein illustrated the transmission effect with a scenario involving NVIDIA: if the company operates at a 75% gross margin and seeks to preserve that margin, any HBM cost increase might need to be multiplied materially when passed on to customers to maintain margin levels. Combined with higher DRAM and NAND prices, Bernstein estimates hyperscalers' data center capital expenditures could be roughly 30% higher to cover increased memory costs.
While Bernstein expects AI investment to continue, it warned that a re-calibration of costs across the supply chain is likely, with weaker suppliers potentially squeezed in the process.
What this means for markets and sectors
These analyst moves collectively underscore two overlapping themes for AI-related markets: memory dynamics are driving fresh, and in some cases aggressive, re-ratings among suppliers; and certain high-growth names are facing valuation scrutiny that tempers immediate upside expectations. Memory suppliers, semiconductor capital equipment firms and GPU makers sit at the center of the market-level implications. At the same time, defensive parts of the AI value chain - data center operators, telecoms and select payments businesses - are attracting reallocations from active managers seeking income stability and lower cyclicality.
Bottom line
The week’s analyst activity reflects a market still in the process of pricing the structural shifts underway in AI hardware and infrastructure. Strong operational beats and multi-year contracts have supporters upgrading memory names; elevated IPO valuations and uneven profitability profiles have others recommending caution. Meanwhile, portfolio managers are rebalancing exposures between fast-moving semiconductor winners and defensive, cash-generative operators as they navigate potential HBM-driven cost inflation and ongoing index-driven volatility.