Morgan Stanley
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 17
- Morgan Stanley lifts Q4 DRAM, NAND contract price forecasts; compute still top AI pick
A Morgan Stanley report raised its memory contract-price forecasts: Q4 2026 DRAM contract prices are now seen up 10–15% quarter over quarter (from +3–8%), with PC DRAM at +13–18% and server DRAM at +10–15%, while NAND moves to +15–20% from +0–5% and enterprise SSDs reach +23–28%. The bank forecasts 2027 HBM average selling prices up 121% year over year, with HBM4 at US$3–5 per Gb and HBM4e above US$5, but says Nvidia's Rubin ramp is behind schedule, with 2026 VR200 NVL72 rack shipments estimated below 10,000 units versus an original target of about 15,000. It keeps its AI supply-chain priority of compute first, networking second and selective memory allocation, and projects DRAM capex of US$98 billion in 2026 (+82%) rising to US$150–160 billion in 2027.
科技新报 · 🔥 10 - Morgan Stanley keeps TeraWulf Overweight, sees 463% upside tied to Anthropic lease
TeraWulf shares extended their October slide on Thursday as rising borrowing costs weighed broadly on AI infrastructure stocks, according to a Barron's report republished by TMTPost. Morgan Stanley analyst Stephen Byrd reiterated an Overweight rating and a $65 price target, implying roughly 463% upside, arguing the selloff does not reflect the true value of the company's data center contracts. The report cites TeraWulf's 20-year lease with Anthropic, expected to generate about $19 billion in contract revenue, and Morgan Stanley's expectation that an additional 336 megawatts of data center capacity will come online by 2027.
钛媒体 · 🔥 5 - Private Tech Investor Enthusiasm Persists as Public AI Debt Worries Grow
At Morgan Stanley’s annual private tech companies conference in Dana Point, Calif., this week, investors remained enthusiastic about startups—especially in the defense and inference sectors—and were impressed by the growth of companies like Fireworks AI, The Information reported. Meanwhile, public-market investors are increasingly worried about overborrowing by AI-related companies, as seen in skyrocketing yields on CoreWeave bonds, a selloff in semiconductor stocks and the yanking of Australian data center operator Firmus’s IPO. Diana Doyle, Morgan Stanley’s head of technology equity and equity-linked capital markets, described a “stark disconnect between the public markets and the private markets.”
The Information · 🔥 5 - CarbonAI raises $12M for AI agents that underwrite commercial real estate loans
CarbonAI has raised $12 million led by Cowboy Ventures, with Timber Grove Ventures, Bling Capital and Max Ventures participating; no stage or valuation was disclosed. The New York company sells only in the US, and its agents pull figures from rent rolls, leases, financial statements and legal documents, build the underwriting model, draft the memo and flag risks, with every number linked back to its source document and judgment calls routed to the lending team. CarbonAI says it has seven-figure contracted revenue within six months and up to 90% time saved on selected underwriting tasks, without giving a methodology; Greystone is the only customer named. The report notes the EU AI Act treats consumer credit decisions as high-risk but not corporate lending, though personal guarantees common in commercial property loans can blur that line.
The Next Web · 🔥 4 - Morgan Stanley CEO: AI capex may hold near $1 trillion a year for a decade
Morgan Stanley CEO Ted Pick said on the bank's third-quarter earnings call that the firm is "double overweight" AI and that investment in AI infrastructure will continue at a massive scale over the next several years, while warning the buildout will hit bottlenecks and periods of overbuilding. Citing Chinese outlet Jiemian via 36Kr, Pick said AI infrastructure spending currently averages about $1 trillion a year, and while growth may not keep accelerating from that level, it could plausibly stay at that scale for another decade. He said the bank is closely watching the market even as it keeps supporting the sector.
The Information · 🔥 0 - Banks move to offload AI data center debt they helped underwrite
According to Financial Times and Bloomberg reporting, JPMorgan, Morgan Stanley, SMBC and MUFG are exploring significant risk transfers and similar deals to shift the riskiest slices of AI data center loans to outside investors as they approach internal limits on exposure to a single sector. In one case, banks including JPMorgan and MUFG have spent more than six months trying to distribute $38 billion in construction debt tied to a Texas and Wisconsin data center project leased to Oracle, with some lenders willing to sell pieces to non-bank buyers at a discount. Oracle's five-year credit default swaps reportedly rose to around 151 basis points in early October, a level last seen in 2009, after S&P cut its rating to BBB- in July, while investors rotate toward data centers leased to investment-grade hyperscalers such as Microsoft and Google.
Hacker News · AI（100+ 分） · 🔥 0 - Tencent, ByteDance turn to debt to fund AI despite large cash reserves
A TMTPost article, citing foreign media reports, says Tencent launched a US$5 billion offshore bond issue on Oct 13, its second overseas fundraising this year after a nearly US$4.7 billion deal in June, while ByteDance earlier in September closed a US$29.6 billion syndicated loan. It notes both remain cash-rich — Tencent reported Q2 net profit of RMB 56.02 billion and about RMB 200 billion in cash, and ByteDance's first-half revenue was reported at US$120 billion with US$20 billion net profit — yet AI capital spending is heavy: Tencent's Q2 free cash flow was negative RMB 13.8 billion with capex of RMB 52.78 billion, and ByteDance reportedly raised its 2026 capex ceiling to US$70 billion. The piece also cites Alibaba's HK$80 billion share placement in August and a Morgan Stanley estimate that global AI-related debt issuance reached about US$450 billion by September, more than double all of last year.
钛媒体 · 🔥 0 - Qwen Go, Muse and Manus: the economics of personal AI agents
A TMTPost analysis compares how three personal AI agents plan to pay for themselves: Alibaba's Qwen Go opened experience applications on Oct 13, with gray-test tasks ranging from car-maintenance quotes to negotiating flight-delay compensation with ticketing support, backed by Alibaba's opening of Taobao and Tmall's 4 billion-plus product catalog plus orders, logistics and after-sales, and April access to China Eastern search, booking, seat selection and check-in. The piece cites an Oct 12 Morgan Stanley model putting Meta's Muse at roughly $445 per user per year against about $70 in average annual revenue per global Meta user, with the $20 subscription tier, at $240 a year, covering only about 54% of cost; Citi projects Muse revenue above $27 billion by 2030, about $23 billion of it transaction-related with commissions over 80%. The author also notes Manus launched Flex on Sept 29, letting users bring their own model API keys and separating inference fees from execution service fees, and cites a 32% runtime-cost reduction for Cascade versus an older system in one test configuration.
钛媒体 · 🔥 0 - Anthropic to spend $100M training 10,000 forward-deployed engineers
Anthropic said it will invest $100 million in a program to train Frontier Deployed Engineers and launch a Claude Frontier Academy, with a goal of training 10,000 people by the end of 2027. Companies nominate engineers rather than individuals applying directly, and candidates must arrive with a defined Claude project; the first cohorts have started in New York and San Francisco — with reports also citing London — drawing participants from Accenture, Bain, Capgemini, Deloitte, McKinsey, Morgan Stanley, Novo Nordisk and Commonwealth Bank of Australia, among others. Participants first take several days of in-person training, practicing use-case selection and agent-system deployment in simulated enterprise settings, then pass a practical assessment before returning to their own companies for a 12-week residency with continued Anthropic mentoring and a final certification. Anthropic executives said strong candidates need familiarity with agentic AI systems, problem-solving and communication skills and solid engineering backgrounds, and singled out curiosity as a key trait, framing the effort as a bid to set a "gold standard" for AI skills training.
IT之家 AI · 🔥 0 - Humanoid robot shipments up 4x, but the industry has no profitable model yet
IDC and Frost & Sullivan both published humanoid robot reports in late September, agreeing the market is scaling but disagreeing sharply on the numbers: IDC counted nearly 25,000 units shipped globally in the first half of 2026, up 432.1% year on year, on $740 million in revenue, while Frost & Sullivan counted 36,000 units and $540 million in revenue, implying an average unit price less than half of IDC's. The article says revenue is concentrated in expensive full-size machines — roughly the priciest fifth of shipments took more than three-quarters of revenue — while deployment in actual production work accounted for under 15%, and no company has shown sustained profit: UBTech, the revenue leader, lost RMB 339 million in the half, whereas Unitree is profitable mainly on volume sales of half-size units such as the G1 (over 5,500 units in 2025, about a 34.8% non-recurring net margin). It also cites Mech-Mind founder Shao Tianlan's public criticism of self-dealing data-collection centers and Galbot's denial and police report, and argues that full-size share, commercialization stage and unit prices are more meaningful indicators for the sector.
36氪 人工智能 · 🔥 0 - Biren plans HK$4.02B placement, total raise hits HK$17.26B
Biren Technology announced on Oct. 8 a placement of 130 million H shares at HK$31.8 apiece to no fewer than six placees, about 4.76% of post-issue share capital, for estimated net proceeds of roughly HK$4.02 billion. The price was a 9.76% discount to the previous close, and the stock fell 11.85% to HK$30.36 that day, closing at HK$31 on Oct. 9 for a market value of about HK$80.6 billion. Including a roughly HK$6.204 billion net raise at its January listing and a HK$7.038 billion placement in July, Biren has raised about HK$17.261 billion (about RMB 14.7 billion) since listing, the most among listed Chinese GPU makers. The company says about 70% of this round goes to strategic supply-chain procurement, production preparation and commercialization for its next-generation flagship BR20X, which it benchmarks against NVIDIA's H200 and expects to enter post-tape-out validation and customer testing in Q4, with large-scale deployment in 2027.
钛媒体 · 🔥 0 - AUO hits 16-year high on NVIDIA glass-substrate talk: three keys to the panel shift
On Sept. 30 Taiwan's panel stocks grabbed the spotlight as AUO and Innolux both closed limit-up, with AUO at NT$38.55 — an roughly 16.5-year high — after a Sept. 24 Korean-media report said NVIDIA is evaluating glass substrates for next-generation AI chip packaging, a claim NVIDIA has not confirmed. Most analysts' price targets for AUO sit at NT$22.5–35.2, meaning the stock has run past most valuation models. The report lays out three keys: Chinese makers BOE, TCL CSOT and HKC are halting lines for 5–7 days over the National Day holiday, cutting utilization from about 82% to the low 70s and issuing price-hike notices that state no amount; AUO is rumored to be selling its L5C and L7A fabs to TSMC for about US$1.2 billion (roughly NT$38 billion), unconfirmed by both sides and complicated by HP and Samsung supply arrangements; and new businesses such as glass substrates, CPO and Micro LED remain in sampling and validation, with Yole projecting the glass-core substrate market at only "hundreds of millions of dollars" by 2030 and chairman Paul Peng saying maturity is two to three years away.
科技新报 · 🔥 0 - Innolight's triple bind as optical modules become a policy flashpoint
An analysis piece lays out three pressures on Chinese optical-module maker Innolight: faster catch-up by peers that shortens each product generation's advantage and pushes prices down; platform players such as Nvidia reshaping system definition through NPO/CPO and locking up upstream optics (it says Nvidia invested $2bn each in Lumentum and Coherent in March 2026); and US market-access rules built on equipment authorisation and covered lists, after Reuters reported in August that the FCC was drafting restrictions and four senators proposed a bill in September targeting sensitive federal systems with a five-year transition. It cites Innolight at roughly 22.4% of the 2025 global optical-module market, with US revenue rising from 57.3% in 2025 to 61.7% in Q1 2026, and H1 2026 revenue of RMB 41.78bn with net profit of RMB 13.65bn.
虎嗅 AI · 🔥 0 - Zhipu's GLM release cycle is turning Chinese LLM makers into 'product-cycle stocks'
An analysis republished by 36Kr (originally from Silicon-Based Observation Pro) argues Chinese large-model companies are increasingly priced like carmakers — 'product-cycle stocks' whose valuations reset with every model release. It cites Zhipu: shares listed at HK$116.2 (about HK$51.8bn market value), peaked at HK$1.07tn on June 22, and fell to HK$696 and roughly HK$324.1bn by October 7, down 76.6% from the high, even as API calls and ARR kept growing. Three of four major model launches this year lifted the stock on the day (average +12.3%), including GLM-5 on February 12 (+28.68%, then +56.22% over five sessions) and open-sourced GLM-5.2 on June 17 (+12.62%); GLM-5.3 on August 14 fell 3.57% after a 40% pre-release run-up.
36氪 人工智能 · 🔥 0 - Humanoid robot hype vs. reality: AI gains won't change daily life soon
Humanoid robots are being sold as AI's next wave: Tesla's Optimus is pitched as a future product that could take on almost all human labor at roughly $20,000 per unit, with Elon Musk predicting public sales by the end of 2027, and Nvidia's Jensen Huang, Marc Andreessen and Morgan Stanley offering similarly bullish forecasts. Many robotics researchers are skeptical, arguing that methods which worked for language and images understate the difficulty of the physical world, where tasks vary endlessly and training data cannot cover every possibility. The article highlights vision-language-action models such as Google DeepMind's Gemini Robotics running on the ALOHA 2 platform, which can perform relatively complex tasks like packing a lunchbox or doing origami but is likely to fail outside its training set; world models are presented as a possible alternative path, attracting backing for startups like World Labs and AMI Labs.
MIT Technology Review AI · 🔥 0 - Anthropic launches Claude Frontier Academy with $100M to train 10,000 engineers
Anthropic announced Claude Frontier Academy, a $100 million program aiming to train 10,000 forward-deployed engineers by the end of 2027, with the first cohort drawn from Accenture, Bain, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk, among others. Training combines a four-day in-person bootcamp with a 12-week real project inside the participant's own organization, leading to Claude Resident Engineer and, after a second assessment, Claude Frontier Deployed Engineer credentials, with the first final certifications expected in early 2027. Enrollment is currently by company nomination through Anthropic's account or partner teams.
36氪 人工智能 · 🔥 0 - Morgan Stanley: Marvell AI infrastructure chip unlikely to threaten Aspeed BMC
At its October 6 Investor Day, Marvell introduced a new custom AI infrastructure management processor based on 5nm OCTEON IP, claiming roughly 1,000x the performance of traditional server BMCs, with projects underway at several hyperscalers. Morgan Stanley said the part is closer to an infrastructure management controller (IMC) than a BMC, carries an ASP of several hundred dollars, and may not ramp in volume until late 2027 or early 2028, so it is unlikely to replace existing BMC designs near term and could instead expand the total addressable market for AI infrastructure management.
科技新报 · 🔥 0
Experience and discussion from the community
Share my Morgan Stanley experienceAsk about Morgan Stanley
Nobody has shared their experience with Morgan Stanley yet.