Anthropic's Historic $30 Billion Round
Research Rundown #172, plus updated reports on Mach Industries and Discord
Research Rundown
Today’s rundown features perspectives on three of the most noteworthy tech stories this week.
Anthropic’s Historic $30 Billion Funding Round
On Thursday, Anthropic closed the second-largest venture funding round in history, a $30 billion Series G at a $380 billion valuation. The round came just five months after Anthropic’s $13 billion Series F at a $183 billion valuation, meaning the company has doubled its valuation in less than half a year. Anthropic has raised a total of $63.7 billion in just five years since its founding in 2021.
The company’s revenue growth has been equally rapid. As Anthropic noted in its blog post announcing the Series G, “it has been less than three years since Anthropic earned its first dollar in revenue”. In January 2023, it had $0 in revenue. By January 2024, this had increased to $100 million, which reached $1 billion by January 2025. This week, the company reached $14 billion in run-rate revenue. It attributed its revenue growth to Anthropic’s “position as the intelligence platform of choice for enterprises and developers.”
This claim seems to be borne out by the data. Last week, we covered the release of Claude Cowork, which was so good that it caused a massive drawdown in the public software market. Meanwhile, Claude Code is rapidly gaining adoption, with SemiAnalysis noting recently that 4% of GitHub public commits are already being authored by Claude Code, and predicting that this would reach 20% by the end of 2026. Meanwhile, Ramp published a report this week which found that, among Ramp’s clients, Anthropic was having a “breakout month”. As Ramp described it, “Anthropic adoption grew from 16.7% to 19.5%, one of its largest monthly gains since we started tracking. Meanwhile, OpenAI adoption slipped from 36.8% to 35.9%, giving back some of last month’s gains.”
AI Safety Concerns Cast a Pall Over Anthropic’s Success
Not all the headlines about Anthropic this week were positive. On Monday, a senior Anthropic safety researcher who led the Safeguards Research Team, Mrinank Sharma, resigned from the company, saying that throughout his two-year tenure at Anthropic, “I’ve repeatedly seen how hard it is to truly let our values govern our actions” and that “the world is in peril” from AI.
This is particularly noteworthy given Anthropic’s founding mission and CEO Dario Amodei’s warnings in recent months on AI safety. Amodei has long been interested in AI safety, having co-authored a paper entitled “Concrete Problems in Safety” in 2016. By December 2020, when he and a group of researchers at OpenAI decided to leave to found Anthropic, he said they were motivated not only to continually scale models, but also to ensure alignment and safety. This remains the company’s public positioning, as it continues to proclaim “we build safer systems” and that “we treat AI safety as a systematic science.”
Amidst the backdrop of Anthropic’s longstanding rhetoric emphasizing AI safety, and following Amodei’s 38-page essay published last month that warned of the dangers posed by accelerating AI progress, as well as his proclamation at Davos along with DeepMind CEO Demis Hassabis that it would “be better for the world” if there was a slowdown in AI progress”, Sharma’s resignation letter seems especially significant. If Sharma’s accusation that the values so vocally espoused by Anthropic’s CEO do not “govern its actions” is accurate, it begs the question of what does.
Memory-Hungry AI Starves Consumers
The global semiconductor industry has been sounding the alarms around a severe supply contraction in memory chips for the last several months. Driven by the allocation of NAND, DRAM, and high bandwidth memory manufacturing capacity to hyper-scale AI clusters, primary fabricators (like SK Hynix and Samsung Electronics) have confirmed that all advanced process nodes are fully booked to meet pre-contracted volume for next-generation GPU interconnects through Q4 2027.
Spot market pricing for standard DRAM chips has surged 600% in this period, triggering margin calls across the consumer electronics supply chain. Automotive and consumer tech manufacturers have been hit hardest, expecting increased prices and lower memory supply for 2026 production cycles. Analysts project that the supply-demand disequilibrium will persist until the operationalization of new fabrication facilities in early 2028.
On its face, the implications for memory-intensive consumer technology are already dire. But the drivers for this surge may be more calculated: are AI companies buying up memory just to leave less for new consumer devices, limiting on-device memory and driving up demand for their massive cloud compute stacks? Or is buying up memory just a backstop to create demand for cloud compute in the case that all the extra capacity isn’t needed for AI? We’ll cover these outcomes in full in our forthcoming deep dive on AI and memory chip supply next week.
Latest Research
Discord filed confidentially for an IPO last month. See our updated report here.
Mach Industries has a new President and CSO. See our updated report here.
What We’re Reading
Anthropic has finalized a $30 billion Series G financing round led by GIC and Coatue at a post-money valuation of $380 billion to accelerate the infrastructure scaling required for its next-generation Claude models.
In conjunction with this capital expansion, Anthropic announced an initiative to fund 100% of necessary grid upgrades and subsidize consumer electricity rate hikes, ensuring its data center growth aligns with the “Power for the People Act” requirements.
OpenAI officially began testing advertisements for the “Go” tier and free users in the United States on February 9, 2026, integrating sponsored content at the bottom of model responses to defray rising computational costs.
OpenAI also submitted a formal memo to the House Select Committee on China alleging that competitor DeepSeek has been “distilling” US-based models to illicitly replicate proprietary capabilities without authorization.
Elon Musk has reorganized xAI into four specialized divisions following the departure of co-founders Jimmy Ba and Tony Wu, consolidating resources from across his business portfolio following the company’s merger with SpaceX.
SpaceX confirmed a delay to its manned Mars timeline to prioritize contractual milestones with NASA, focusing all immediate efforts on a lunar landing by March 2027.
Microsoft shared that it is developing an internal 500-billion-parameter model, MAI-1, led by Mustafa Suleyman, to reduce its long-term dependence on OpenAI’s GPT architecture for core products.
Meta plans to deploy a “Name Tag” feature for its Ray-Ban smart glasses in 2026 that utilizes facial recognition to identify strangers in real-time, a move that has ignited significant debate regarding privacy and “super-sensing” technology.
Spotify reported that its lead engineers have utilized an internal AI tool named “Honk,” powered by Anthropic’s Claude Code, to ship over 50 new features since December 2025 without manually authoring a single line of code.
Erebor Bank, the venture founded by Palmer Luckey to service the cryptocurrency and defense sectors, has successfully obtained a national bank charter, marking the first such approval under the current administration.
At Contrary Research, we’ve built the best starting place to understand private tech companies. If you're interested in researching and writing about tech, apply here for our Research Fellowship!
At Contrary Research, we’ve built the best starting place to understand private tech companies. If you're interested in researching and writing about tech, apply here for our Research Fellowship!