New Street sets $105 Anthropic target ahead of IPO | Magica
New Street sets $105 Anthropic target as compute costs test IPO case
Editorial Team
••📖7 min read
New Street Research has initiated Anthropic at Buy ahead of its planned IPO. Its $105 target bets on expanding AI demand, while the filing shows surging revenue and improved customer-serving economics alongside volatile operating results and vast infrastructure commitments.
New Street Research's $105 target is about 35% above the top of Anthropic's proposed $70–$78 IPO range.
The filing implies an 85.5% gross margin for the first nine months of 2026; third-quarter operations still made a loss.
A concurrent $26 billion private placement would add financing from technology companies, including a major competitor.
New Street Research, an investment research firm, has initiated Anthropic at Buy ahead of its planned stock-market debut. The $105 target backs the developer of Claude, a family of AI models and tools for software development and other business tasks, as a beneficiary of expanding spending on knowledge work. Anthropic's preliminary filing shows both the demand behind that argument and the expense of supplying it.
Pierre Ferragu, the analyst behind New Street’s Anthropic Buy initiation. Photo: New Street Research. Source: New Street Research.
The analyst's market forecast leaves room for rivals
Pierre Ferragu, who heads New Street's global technology infrastructure and Software & AI Stack research teams, previously spent more than ten years at Bernstein covering sectors including networking, cybersecurity and semiconductors. His employer biography lists the Anthropic initiation on November 16. In an account quoting his research note, Ferragu calls Anthropic a “generational IPO” and the best-positioned provider at the frontier of AI today.
His analogy is cloud computing. A report on the initiation describes his comparison with cloud capturing 30% of incremental IT spending between 2010 and 2025. Ferragu expects AI to spread twice as quickly across an addressable market ten times larger, reaching nearly 7% of knowledge-work spending in 2030 and capturing about 25% of incremental spending through 2040.
The Alphabet-backed drug developer has announced $7.1 billion in funding in six months. Its new capital will expand proprietary AI models and clinical infrastructure as company-reported preclinical progress awaits testing in patients.
Editorial Team
Those forecasts concern AI's overall market. Anthropic's share and profitability require a separate judgment. Ferragu himself argues that both frontier and open-weight models have sustainable business models and broad opportunities; his thesis accommodates competing suppliers.
Anthropic's November 10 preliminary prospectus names Google and OpenAI as primary frontier-model competitors. It also identifies alternatives for workloads that prioritize cost over intelligence, including models whose weights are available for others to use. Larger competitors can bundle AI into existing products, subsidize prices and use established distribution relationships. These are company-disclosed risks, rather than proof that a rival has taken market share.
Company-reported quarterly revenue and compute and infrastructure expenses. Compute includes research and model development as well as customer service; the gap is not gross profit. Chart: Magica; source: Anthropic. Source: Anthropic preliminary prospectus.
A short commercial history, with demand already at scale
The prospectus dates Claude's commercial launch to the first quarter of 2023. By September 30, 2026, Anthropic reported more than 11,900 customers generating at least $100,000 each in trailing-12-month revenue, including more than 1,900 generating at least $1 million. Those overlapping groups indicate the scale of business spending already attracted to its products.
The filing's unaudited quarterly results show revenue rising from $1.382 billion in the third quarter of 2025 to $16.956 billion a year later—more than twelvefold. The latest sequential comparison is less reassuring on costs:
$ billions, rounded
Quarter ended June 30, 2026
Quarter ended September 30, 2026
Revenue
11.56
16.96
Compute and infrastructure expenses
5.56
9.86
Operating income (loss)
0.97
(0.57)
Calculated from the filing's unrounded figures, revenue increased about 47% quarter over quarter and compute and infrastructure expenses about 77%. The latter rose from 48% to 58% of revenue. This expense category includes research and model development as well as serving customers, so the comparison does not establish a customer-service gross margin.
The filing's unaudited supplemental breakdown by function provides a stronger measure of customer-serving economics. For the nine months ended September 30, 2026, cost of revenue was $4.812 billion against revenue of $33.242 billion. Subtracting that cost from revenue and dividing by revenue implies a gross margin of about 85.5%, up from 29.4% in the same nine months of 2025. Cost of revenue primarily covers customer-serving compute, plus certain personnel and overhead costs.
That improvement supports the bullish case, but excludes the continuing expense of developing the models customers buy. Research and development expenses were $19.24 billion in the first nine months of 2026, including compute for research, model development and internal use. High gross margins can coexist with substantial investment needed to maintain frontier capabilities.
Anthropic attributes the return to an operating loss partly to $5 billion of compute under short-term commitments, which it says currently costs significantly more than comparable capacity under multiyear contracts. Another major factor was $2 billion of noncash charges tied to its donation-matching program. The loss therefore cannot be attributed entirely to compute.
Pricing also matters. Anthropic says reductions introduced with new model versions and contractual discounts offset approximately $600 million of the revenue increase from higher usage between the second and third quarters. Growing usage alone does not determine how much revenue each new unit of capacity earns.
The IPO finances capacity while preserving founder and trust control
The proposed offering comprises 825 million shares: about 801.35 million sold by Anthropic and 23.65 million by existing shareholders. At $78, gross proceeds across both groups would be $64.35 billion before any underwriter option. Anthropic receives none of the selling shareholders' proceeds.
The financing extends beyond the public offering. Seven technology companies, including Google, Microsoft and Nvidia, have agreed to a concurrent $26 billion private placement of nonvoting Class N shares at the IPO price. It is conditional on completion of the offering and other requirements. Google's portion funds part of an existing investment commitment; Google is also one of the competitors Anthropic identifies.
At the $74 midpoint, Anthropic estimates combined net proceeds to the company from the IPO and private placement of $84.6 billion, after estimated fees and expenses and before any underwriter option. Its intended uses include compute and infrastructure, working capital, and hiring and retention.
That money would supplement substantial existing resources. At September 30, Anthropic reported $95.6 billion in cash and equivalents and $36.3 billion in short-term investments. Operating activities generated $6.6 billion of cash in the first nine months of 2026, despite an operating loss over that period. Accounting losses, cash generation and financing needs describe different parts of the business.
Public investors would also acquire limited influence over its direction. As a Delaware public benefit corporation, Anthropic must balance shareholder interests, those materially affected by its conduct, and its mission of responsible AI development. A founder-controlled entity will hold 50.1% of general voting power. Separately, the Long-Term Benefit Trust, intended as an independent body of financially disinterested trustees, generally has authority to elect and remove directors constituting a board majority.
Anthropic’s proposed IPO range is $70–$78 per share; New Street’s $105 target is an analyst judgment. Chart: Magica; sources: Anthropic and New Street Research. Source: Anthropic preliminary prospectus.
What pricing and the next results must resolve
Ferragu expects AI growth to remain supply constrained until 2030. Anthropic says it spreads workloads across multiple suppliers and chip types to optimize price and performance, but warns that chip supply, electricity availability and data-center construction can constrain capacity. Securing financing does not itself deliver usable compute.
The filing discloses approximately $900.9 billion of purchase and lease commitments as of and subsequent to September 30, 2026, including short-term arrangements and leases not yet commenced. Substantially all relate to compute and infrastructure. About $70.7 billion is due in the following twelve months, and $451.6 billion is expected before December 31, 2030; substantially all the remainder extends through 2045.
The total has qualifications. About $186.2 billion relates to arrangements subject to exercisable termination rights or conditions before commencement. Separately, the filing estimates another $150 billion–$175 billion in payments under arrangements without fixed dollar commitments, with the amount dependent on deployment and other conditions. These figures describe obligations and estimates across future periods, rather than current annual expenses.
The immediate decision is the entry price. Pricing is expected on November 18 and Nasdaq trading on November 19, according to the report on the initiation. Dividing $105 by the proposed $70–$78 endpoints and subtracting one gives potential upside of 50% to about 35%. A higher trading price would reduce that gap; the analyst's target is a judgment, rather than an offering term or promised return.
After pricing, the next results need to show whether Anthropic can maintain its customer-serving margins while funding model development and using newly delivered capacity efficiently. Customer retention, realized prices and operating results will test whether it can convert the broad market opportunity into returns for its new shareholders.
Microsoft enters Ignite with a previously announced Copilot redesign that combines office applications and autonomous agents. Availability, usage-based charges and reliability will determine what that expansion means for customers, while October's missing paid-seat update leaves adoption harder to assess.
A proposed Equity–Pact agreement would give UK film and TV performers consent and payment rights over digital replicas. Members must approve the package, which leaves explicit protections against AI training unresolved.
The French cloud provider plans to release Olala mini 7A1B on November 19, backed by acquired expertise and European public computing support. Its first deployment offers cost claims, but no published benchmarks yet establish its competitive quality.
US stock futures fell before Nvidia's first earnings report since Jensen Huang's departure, with cash generation, tighter export controls and rate-hike expectations clouding the AI trade ahead of Anthropic's expected market debut.
Nvidia has created a standing board committee to oversee national security and export compliance after co-founder Jensen Huang’s departure. Investigation findings and operational changes remain unresolved.
Microsoft's CEO says team sizes, workflows and skills will change as AI reshapes engineering. About 60,000 employees are eligible for voluntary exits, but uptake and the prospect of later layoffs remain unresolved.
Hong Kong officials say financial firms and their human overseers remain responsible for AI agents. With 36 projects selected for sandbox testing, the unresolved task is turning that principle into reliable identity checks, delegated authority and human intervention.
Meta says stronger AI checks and India-specific tuning keep more suggestive and violent content out of Instagram recommendations. Its reporting metric leaves exposure unmeasured, while a government request to change ranking systems drew objections over the ministry’s authority.
OpenAI says NATO has used up to 300 billion raw input and output tokens a month for cyber defence. Access through NATO’s technology agency and an AI-agent exercise shows uptake, while costs, comparative performance and NATO-specific security results remain undisclosed.
Numa’s Series C finances a move from customer conversations into sales and service operations. Automaker distribution gives it reach, but vendor approvals and company-reported savings leave the broader operating-system pitch to prove.
Google’s research tool can turn uploaded sources into interactive dashboards, study aids and trackers. Shared links require full notebook access, personal inputs stay separate, and generation draws on compute-based usage limits.
Records detail Chinese data-center investments linked to Hexi founder Kang Wei, husband of former Megaspeed chief Alice Huang. Filings document financing and Huawei business, but do not establish a route for restricted Nvidia chips.
ElevenLabs plans to acquire Kotoba, the startup behind Japanese speech technology used in Salesforce’s Agentforce Voice. The deal adds a 15-person team and deployed products to its existing Japanese operation, but financial terms and integration plans remain undisclosed.
Tesla plans €3 billion of expansion at its profitable Grünheide factory to build Cybercabs from 2029, eventually targeting up to one million annually. The driverless taxi still lacks German and EU approval, and its capacity target differs from an older plan for the whole site.
Our World in Data’s comparison finds negative views of AI exceed positive ones across five probability-based surveys. The primary results also reveal substantial mixed views, acceptance of some applications and a sharper divide over who controls development and bears its local costs.
NVIDIA's Nemotron for Science pairs an existing open model with nearly 50,000 chemistry questions and scientific post-training recipes. Partner results suggest scope for specialized agents, but the reported accuracy and inference savings do not establish the full cost of adapting them.
AMD says the Lux AI supercomputer is open to scientific users at Oak Ridge. Half its annual capacity supports the Genesis Mission and half is reserved for paid access, including AMD development, while workload limits and undisclosed rates shape its practical value.
Apple is adding app-by-app and folder-specific permissions, repeat approval requests and clearer warnings in macOS 27.3 beta 2. Existing permissions remain intact, limiting the immediate reach of the privacy changes.
OpenAI reportedly discussed buying Goodfire for more than $5 billion, over four times its February financing valuation. No deal resulted. A disclosed paid assessment agreement offers cooperation without ownership, while the technology’s safety benefits remain to be demonstrated.