Investing.com — Here are the biggest analyst moves in the area of artificial intelligence (AI) for this week.
Susquehanna hikes Micron target to $2,000 after record-breaking earnings
Susquehanna has raised its price target on Micron to $2,000 from $1,750 after the memory chipmaker posted record-breaking fiscal third-quarter results and guidance that comfortably exceeded consensus.
Micron reported revenue of $41.46 billion and adjusted EPS of $25.11, well above consensus of $35.91 billion and $20.86, with gross margin of 84.9% also surpassing expectations of 80.8%.
For the current quarter, the company guided for revenue of around $50 billion and EPS of approximately $31, again ahead of Street expectations.
Central to the upgrade are Micron’s 16 signed strategic customer agreements (SCAs) — non-cancellable, take-or-pay contracts typically running five years — which Susquehanna sees as a structural shift in how the company generates earnings. The agreements cover roughly one-fifth of DRAM bits and one-third of NAND bits, and when fully executed, Micron expects more than half of total revenues to fall under such contracts.
Importantly, the deals include price floors that lock in gross margins well above any prior-cycle peak, even in a downturn scenario. Susquehanna now expects Micron to generate free cash flow in excess of $110 billion in fiscal 2027 (FY27), the bulk of which should flow to shareholders.
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, with data center SSD revenue exceeding $5 billion in the quarter alone.
Micron now sees the HBM total addressable market crossing $100 billion in 2027, a year earlier than previously expected.
Susquehanna, which already had estimates well above consensus, raised them further and now projects annualized EPS potentially reaching $200 as Micron exits FY27.
Argus initiates SpaceX at Hold, says IPO valuation will take years to grow into
This week, brokerage firm Argus has initiated coverage of newly-listed SpaceX with a Hold rating, arguing that the company’s record-breaking IPO valuation will take years to grow into despite strong top-line momentum.
SpaceX completed its IPO earlier this month, raising total proceeds of $85.7 billion at a valuation above $1.75 trillion — the largest IPO in history — implying a price-to-sales multiple of approximately 95 times 2025 revenues.
Analyst Steve Silver flagged the valuation as a key concern, noting that SpaceX “is not yet consistently profitable, as its operating plan combines characteristics of both a mature infrastructure business and a venture-style growth investment.” He believes that “it will likely be years before SPCX’s multiples land at more normal levels.”
Since the IPO, shares have been volatile, rising as much as 67% above the offering price before pulling back to trade around 10% above it. Argus expects volatility to persist “given a tight share supply and its early inclusion in many major equity indices,” alongside looming lockup share expirations.
Still, Argus left the door open to an upgrade, saying it “may look to upgrade the stock if it falls sharply on nonfundamental factors or if revenues and earnings accelerate at a faster-than-expected pace.”
Morgan Stanley upgrades Qualcomm after $5bn data center target
Morgan Stanley has upgraded Qualcomm to Equal-weight from Underweight, raising its price target to $231 from $146, after the chipmaker’s investor day revealed a $5 billion data center revenue forecast for fiscal 2027 that the bank said “clearly pushes them into the AI beneficiaries category.”
Analyst Joseph Moore acknowledged the bank had been wrong to remain skeptical of Qualcomm’s AI ambitions. “We have been wrong to be skeptical,” he wrote, adding that management’s $5 billion revenue forecast for next year is “at least 2x higher than expected.”
The analyst was more cautious on Qualcomm’s longer-term target of $15 billion in data center revenue by fiscal 2029, describing it as “more aspirational.” He also flagged execution risk around the accelerator opportunity and server CPUs, noting that a mid-2028 entry into the CPU market “may not meet with the reception that it would have today” given rapidly expanding supply from NVIDIA, AMD, Intel and cloud providers’ custom silicon.
Moore stopped short of an Overweight rating, citing better value elsewhere. “We simply see better value from the dominant players vs more speculative new entrants, and we remain mindful of smartphone headwinds,” he wrote.
Despite lingering questions on the product roadmap, he said Qualcomm’s credible management team and year-to-date underperformance relative to other AI winners made the upgrade difficult to resist, particularly in “the current narrative driven market.”
UBS trims semiconductor weighting after rally, rotates into defensive AI names
UBS has reduced its combined semiconductor and hardware weighting in its AI investment strategy to around 61% from roughly 76%, taking profits following a Micron-driven rally that lifted the SOXX index 4% in a single session and 10% across June.
Profit-taking was concentrated in small- and mid-cap AI supply chain names across optics, baseboard management controllers, substrates, cooling, advanced packaging and analog components.
Still, UBS said its strategy “still retains an overweight of roughly 20-25 percentage points” relative to the Nasdaq 100’s approximately 42% AI semiconductor and hardware exposure.
In parallel, the bank increased exposure to more defensive parts of the AI ecosystem to around 20% from less than 1% in early June, focusing on data center operators, telcos and select payments names with “prudent balance sheets and stable dividend profiles.”
It said “the medium-term AI demand story remains intact, supported by cloud acceleration and agentic AI workloads,” but flagged the risk that a significant decline in hyperscaler stocks “may pressure management to reduce capex commitments in the future.”
Within semiconductors, UBS continues to favor foundry exposure such as TSMC, semicaps including Applied Materials, and memory names such as SK Hynix. The bank maintained a significant underweight to the Magnificent Seven, with a combined portfolio weight of approximately 18%.
Bernstein raises memory trio targets as HBM price surge set to inflate AI capex
Earlier in the week, Bernstein raised its price targets on memory chip giants Samsung, SK Hynix and Micron, forecasting a sharp rise in HBM prices that it expects to drive significant earnings revisions across the sector, while warning that surging memory costs could inflate AI infrastructure spending by as much as 30%.
The broker lifted its target on Samsung to KRW440,000 from KRW225,000, SK Hynix to KRW3,300,000 from KRW1,150,000, and Micron to $1,300 from $510, maintaining Outperform ratings on all three.
It kept its Underperform on KIOXIA, which has no HBM exposure.
The core argument is that HBM prices need to rise substantially to close a widening profitability gap with conventional DRAM. Since Q3 2025, conventional DRAM prices have surged roughly 4.5 times while HBM prices, locked in by annual contracts, have barely moved.
As a result, Bernstein estimates that deploying capacity to conventional DRAM currently generates over twice the revenue and nearly three times the gross profit per wafer compared to HBM.
“This is why memory suppliers and GPU/XPU companies are negotiating 2027 HBM price now to narrow the gap,” analyst Mark Li said.
Bernstein models a 2 to 2.5 times increase in HBM prices next year, somewhat below what would be needed to fully close the gap, on the assumption that suppliers recognize the strategic importance of keeping HBM affordable for the AI ecosystem.
The inflation risk for AI capex stems from how HBM price increases flow through the supply chain. Unlike conventional DRAM and NAND, HBM is packaged inside GPUs and forms part of the cost base of chip suppliers like Nvidia.
If Nvidia operates at 75% gross margin and wants to preserve that margin, it would need to mark up any HBM price increase by a factor of four when passing costs on to customers. Combined with higher conventional DRAM and NAND prices, Li estimates hyperscalers could face data center capex “higher by 30% just to cover higher memory costs.”
The analyst expects AI investment to continue but says “a re-calibration is inevitable,” with costs potentially redistributed across the supply chain and weaker suppliers squeezed in the process.
Related articles
5 big analyst AI moves: Micron price targets hiked, cautious on SpaceX valuation
As Claude disrupts stock market, Anthropic researcher warns ‘world is in peril’
Wolfe Research outlines eight risks that could spark stock declines in 2026