E240: xAI $230B valuation, Databricks $134B valuation; Revolut $75B tender; Wealthfront IPO; raises from Physical Intelligence, Black Forest Labs, Moonshot AI, others

0:00Aaron Dillon here, Managing Director of AG Dillon & Co. Here's the pre-IPO stock update for December 5, 2025. Databricks, a private data infrastructure and AI platform, is raising a new$5 billion round at a$134 billion valuation, equal to roughly 32x its 2025 expected sales of about$4.1 billion. And that's up from 24x sales in its last year's round and 26x the year before that. The company projected 55 % revenue growth this year and has lifted guidance at least twice, increasing its forecast from$3.8 billion to$4 billion and then slightly above that, extending a multi-year streak of greater than 50 % annual growth that is rare among mature software peers.

0:45Databricks' gross margin has fallen more sharply than planned to about 74 % versus a prior internal target of 77 % and down from 79 % a year ago. As AI product usage drives higher cloud storage and GPU costs, leaving the company roughly break even with about$10 million in free cash flow in 2025 after burning, quote, hundreds of millions annually as recently as 2023. So that's pretty good. Around 25 % of revenue now comes from AI products layered on top of its core data lake platform. The company reports more than 20 ,000 customers that include government agencies, the NBA, AT &T, and Shell, with its top 10 customers accounting for less than 15 % of total usage, which implies a limited concentration risk.

1:31The new round will be led by existing investor Insight Partners, which is expected to invest at least$500 million. Then the proceeds will fund large scale employee liquidity and tax related obligations. Databricks valuation sits between infrastructure rivals Snowflake and Datadog, which traded about 21x and 16x forward sales and high multiple Palantir, which trades at 90 over 90x expected sales, placing Databricks in the upper tier of software valuations, but still below most speculative public comps. The company has committed$100 million of spend over several years on open AI models, and Databricks is benefiting today from enterprises racing to automate workflows like HR and IT service management with AI agents built on its platform.

2:20Revolut, a UK-founded digital bank, has completed a secondary share sale that values the company at$75 billion, up from$48 billion post-money valuation in August of 2025, and supported by a capital base of$2.89 billion of primary venture funding to date. The latest transaction was led by KOTU, Greeno, Dragoneer, and Fidelity with participation from NVIDIA's InVentures, Andreessen Horowitz, Franklin Templeton, and T. Rowe Price, and enabled employees to sell stock for liquidity, marking the fifth internal share sale since inception and underscoring one of the more liquid equity programs in private fintechs.

2:55Revolut generated$4 billion of revenue in 2024, up 72 % year over year, while profit before tax rose 149 % to$1.4 billion, and net profit reached about$1 billion. And the management says its business unit, Revolut Business, has now scaled to$1 billion of annualized revenue. The group reports more than 65 million customers globally and has set an ambition to reach 100 million customers by mid-2027 and to expand into more than 30 new markets by 2030, with current operations spanning the EU, the UK, the US, Australia, Japan, New Zealand, Singapore, and Brazil, plus recent or upcoming launches in India, Mexico, and Colombia, and early plans for Argentina and South Africa, and an in-principle payments license in the UAE.

3:41It's products that now covers multi-currency accounts, payments and transfers, crypto insurance and wealth management, including a crypto exchange platform, RevolutX, whose revenue surged 298 % in 2024 to$647 million from$158 million in 2023. The new$75 billion valuation places Revolut among Europe's most valuable private tech firms and implies a revenue multiple of roughly 18.8x on 2024 revenue. with growth and profitability metrics that position it as a rare-scaled neobank, delivering both rapid international expansion and double-digit return characteristics. Revolut has an$81 billion secondary market valuation.

4:25That's up 8.3 % versus this just-announced tender round. Physical intelligence, a one-year-old robotics AI startup known as Pi, P-I for short, has raised a new round of about$600 million at a$5 billion valuation, including the new capital after previously raising roughly 470 million in a 2.4 billion valuation pushing total funding close to 1.1 billion in roughly a year the latest round is led by alphabets capital g with participation from existing investors lux capital thrive capital and jeff bezos with new investors index ventures and t-roll price signaling strong crossover interest and foundational robotic software pi develops general purpose ai models that can control many types of robots for tasks from folding laundry and assembling boxes to bagging groceries and taking toast out of a toaster.

5:17And its latest model, Pi 0.6, learned partly from real-world deployment on physical robots, improving speed and reliability versus prior versions. And earlier funding showed the company had raised$400 million at a$2 billion valuation, indicating a more than 2.5x step up to the current$5 billion mark in under a year as investors bet on a horizontal software layer for robotics. The company's founding team includes ex-Google researchers, Stanford and UC Berkeley professors and early Stripe employee and investor Lockheed Groom, positioning Pi among a cohort of well-capitalized robotic platforms that include Vicarious, Universal Robots, Seagrid and Covariant.

6:02Elon Musk predicts that there's going to be as many as 10 billion humanoid robots on the planet by 2040 at$20 ,000 to$25 ,000 per unit. So Pi's strategy is to build a universal software stack. And Pi Zero and its successors, that's their software program, can run on any robot. So customers do not need bespoke control software for each hardware type, a positioning that could give it leverage across industrial logistics consumer service robot segments if adoption scales. Black Forest Labs, a German foundational model startup focused on image generation editing, has raised$300 million in a Series B round that values the company at$3.25 billion, less than two years after its launch in August of 2024.

6:45The round is co-led by Salesforce Ventures and AMP with a heavy roster of strategic and venture backers, including A16Z, NVIDIA, General Catalyst, Temasek, Bain, Canva, and Figma Ventures. And the proceeds are earmarked for research and development. Black Forest Labs gained prominence when it emerged that Elon Musk's Brock chatbot, that's XAI, used its model for image generation. And its foundation models now power or integrate with tools from Adobe, Pixar, Eleven Labs, and Vercel, giving its broad distribution across creative consumer and developer platforms. The company's latest model, Flux 2, improves both text and image rendering and can use up to 10 reference images to preserve style and tone for new generations, while outputting images at resolutions of up to 4K, targeting both professional and prosumer design workflows.

7:39BlackFour's founding team previously played key roles in the development of stable diffusion at Stability AI. Wealthfront, a digital wealth management platform, has filed an S1 for an IPO on NASDAQ under ticker WLTH and is targeting a valuation of up to$2.05 billion above the$1.4 billion price implied by a proposed but terminated UBS acquisition in 2022. The company plans to raise up to$485 million by selling 34 million shares at$12,$14 per share, of which$21 million will be new primary shares and$13 million will be secondary shares sold by existing shareholders with no proceeds from the secondary component going to the company.

8:22Underwriters have a 30-day option to buy up to$5 million additional shares, which could increase gross proceeds meaningfully if exercised. exercised. Lead bookrunners are Goldman Sachs and JP Morgan with Citigroup, Wells Fargo, RBC as active bookrunners and Citizens, KBW, Stiefel and KeyBank as co-managers, underscoring a full Wall Street syndicate. Wealthfront markets itself as an AI-enabled fintech that automates portfolio management, cash accounts, low-cost loans, financial planning for millennial and Gen Z clients and expects the benefit from strong investor appetite for AI-adjusted fintechs, a trend illustrated by robust demand for listings like Klarna, Chime, and eToro.

9:04Funds managed by BlackRock and Wellington have indicated interest of up to 150 million of IPO shares, a cornerstone that could de-risk the book build. Moonshot AI, a Beijing-based large AI model startup focused on long-form text analysis and reasoning is nearing a new dollar denominated funding round that could value it at over$4 billion up from about a reported or excuse me up from a reported series B round of more than 300 million raised in August last year from investors including Tencent and Gowrong Capital. The new deal is expected to total several hundred million dollars and includes talks with global investor IDG Capital and existing shareholder Tencent as moonshot signals an ambition to pursue an IPO in the second half of next year.

9:54Moonshot is also attracted backing from Alibaba Group and Hong Shan, the rebanded Sequoia China, positioning it among a cluster of heavily funded Chinese foundation model firms such as Minimax and Zipu AI that are now preparing and pursuing listings in Hong Kong and onshore China. The company's open source Kimi K2 thinking model released this month and trained on NVIDIA H800 chips that comply with U.S. export rules tightened in late 2023, ranks among the top models on benchmarking site LM Arena and shows enhanced coding skill and long context reasoning. And Moonshot has signaled that the next generation K3 model is already in the pipeline.

10:41Chinese AI startups still trade at a fraction of U.S. peers, yet the domestic ecosystem has accelerated since low-cost contender DeepSeek emerged in January with models that rival Western quality at far lower prices. xenergy a u.s based nuclear startup focused on small modular reactors that's smrs have raised 700 million in a series d round less than a year after expanding its series c from 500 million to 700 million bring total capital raised in roughly a year to 1.4 billion the total historical funding to about 1.8 billion the new financing led by jane street alongside investors including Ares, ARK Invest, and Point72, will fund the build-out of a supply chain for its XE100 high-temperature gas-cooled reactors.

11:29XEnergy says it has orders for 144 SMRs capable of generating 11 gigawatts of electricity in aggregate, applying about 76 megawatts per unit. Its customer pipeline includes Amazon, Dow, and UK Utility, Centrica. Amazon's climate pledge fund led the original Series C, and Amazon has separately announced plans to purchase over 600 megawatts of nuclear capacity in the Pacific Northwest and Virginia with up to five gigawatts of X Energy capacity potentially deployed by 2039, reflecting growing data center and cloud demand for firm carbon-free power. X Energy's reactors use billiard ball-sized carbon-coated fuel pebbles with embedded uranium particles and helium coolant to transfer heat to steam turbines, a design used by Japan and China and marketed as inherently safer and more flexible than traditional large reactors.

12:23The company is part of a broader SMR race in which many designs remain unproven at scale, and only a handful of SMR plants exist globally with none operating in the U.S. Yet investor and corporate interest has surged as hyperscalers and industrials seek long-duration dispatchable power to support AI compute growth. marvel technology has agreed to acquire celestial ai a private optical interconnect startup for at least 3.25 billion in cash and stock with the purchase price rising to as much as 5.5 billion if celestial meets revenue milestones that include two billion dollars of accumulative revenue by the end of fiscal 2029 celestial reportedly valued at 2.5 billion in a funding around earlier this year and backed by investors like Intel board member Lip Boo Tan builds a phototonic fabric for high-speed optical connections among AI accelerators and custom XPU chips.

13:21A critical bottleneck as frontier large language models require systems that link dozens or hundreds of chips. Marvel expects data center revenue to grow to 25 % next year and see Celestial's technology as a way to expand its connectivity, TAM, and deepen wallet share with hyperscalers spending hundreds of billions on AI infrastructure. Today, many AI systems still rely on copper connections and limited bandwidth and cable length, and optical links can deliver higher data throughput and longer reach, but at higher cost. Marvel plans to integrate Celestial's optical engines directly into custom chips and switches for large AI clusters.

14:00The company said the first deployment will target systems built around large custom AI XPUs and that Celestial's technology will be woven into Marvel's existing portfolio for AI and cloud customers. with Amazon Web Services executives publicly backing the deal as an accelerator for optical scale innovation. The transaction expected to close early next year represents one of the larger strategic bets to date on AI networking. Canva, an Australian graphic design platform, is using AI enhanced design tools and newly acquired Affinity Suite to gear up for a public listing that management describes as imminent in the next couple of years.

14:44The company recently hired former Zoom CFO Kelly Shekelberg as a senior finance executive, signaling IPO preparation as it builds out the finance and governance bench. Canva's strategy centers on AI-driven content creation and collaboration tools that aim to replace or augment legacy design workflows for small businesses, enterprises, and individual creators to deliver strong user-driven growth, which in turn supports talent attraction and retention at scale. Canva has a$47.1 billion secondary market valuation. It's up 12.2 % versus its August 2025 round. Sierra, a 21-month-old enterprise AI startup that builds custom service agents, has reached a$100 million annual recurring revenue run rate and is valued at$10 billion after a$350 million round led by Green Oaks, implying a 100x ARR valuation multiple that assumes sustained hypergrowth.

15:42The company serves a mix of tech native and traditional customers, including Deliveroo, Discord, Ramp, Rivian, SoFi, and Tubi on the tech side, plus ADT, Bissell, Vans, Cigna, and Sirius XM outside of tech, and positions its agents to handle tasks like healthcare, patient authentication, processing returns, replacing credit cards, and supporting mortgage applications. Sierra uses an outcomes-based pricing model that charges for completed work rather than flat subscriptions, which may align incentives more closely with customers. The founders, Brett Taylor and Clay Baver, bring deep tech, big tech credentials.

16:22Taylor co-created Google Maps, founded FriendFeed, helped build Facebook Like Button, and sold his productivity startup Quip to Salesforce for$750 million and later served as Salesforce co-CEO. While Baber spent 18 years at Google, leading core products like Gmail and Google Drive. Sierra completes with other AI customer service providers, such as Decagon and Intercom's AI offerings, but currently claims category leadership. And the current$10 billion valuation places it among the most richly priced early stage enterprise SaaS companies on record with investors effectively pricing in a path to multi-hundred million dollar ARR in the medium term.

17:11K2, a satellite manufacturing startup founded in 2022 and staffed with former SpaceX engineers, is raising a new round that would roughly quadruple its valuation to$3 billion, up from$690 million a year ago and a round led by Ultima Capital and Lightspeed, according to corporate filings. The new funding led by venture firm Redpoint Ventures will add to K2's total capital raise, which will reach roughly$475 million after the company issues about$270 million of new shares based on a Delaware charter filing. K2 is based in Torrance, California, and is developing high-powered satellites for higher orbits than low-Earth orbit constellations such as SpaceX and Starlink, with each satellite designed to deliver about 20 kilowatts of power, around four to five times the output of the highest power satellites previously flown, according to co-founder and CTO Neil Kuhnjer.

18:03The company builds satellites largely in-house for about$15 million per unit versus hundreds of millions of dollars for traditional manufacturers like Lockheed Martin and has attracted early interest from commercial customers. In September, it announced a partnership with Swiss operator SES to support an internet network mission starting early next year. And the round continues a trend of multi-billion dollar valuation for early stage space startups founded by alumni of SpaceX and Blue Origin, including Apex Space, Impulse Space and Stokes Space. Anthropic, a frontier AI large language model lab founded in 2021 by former OpenAI leaders and recently valued around$350 billion after a new multi-billion dollar investment from Microsoft and NVIDIA, has launched Claude Opus 4.5, its most capable model for coding and complex enterprise tasks and capstone of its 4.5 model family after Sonnet 4.5 and Haiku 4.5.

19:03Opus 4.5 is positioned for professional developers and knowledge workers, such as financial analysts, consultants, and accountants, and is meaningfully better than prior versions at spreadsheet work, slide creation, and deep research and is state-of-the-art on a wide set of benchmarks, including coding, tool use, and reasoning. On SWE Bench Verified, the model is the first to score above 80%. And according to Anthropic, it also outperforms Google's Gemini 3 Pro and OpenAI's GPT51 on agentic coding benchmarks. Anthropic reports more than 300 ,000 business customers using its models and internal tests show Opus 4.5 scoring higher than any human candidate on the company's take-home performance engineering exam, a sign that the model can't handle complex multi-hour engineering assignments with minimal human assistance.

19:59Opus 4.5 will serve as the default model for Anthropics Pro Max Enterprise plans and is also being tightly integrated into productivity tools through Cloud for Chrome and Cloud for Excel, which are now broadly available to Mac's team and enterprise users after pilot phases. Release also introduces memory and long context improvements that enable endless chat by compressing context on the fly without user interruption and by improving the model's ability to remember the right details rather than simply storing longer stories or critical capability for agentic workflows where an Opus-powered lead agent coordinates haiku-based sub-agents across large code bases and long documents.

20:43These updates arrive in a competitive window in which OpenAI has launched GPT 5.1 and Google has unveiled Gemini 3, and Anthropic is racing to differentiate on enterprise reliability, coding automation, and deep integration into tools like Excel. Anthropic is in advanced talks to acquire Bun, a developer tool startup that streamlines installing, building, and running and testing JavaScript and TypeScript code in what would be Anthropic's first true technology acquisition after a history of talent-only acquihires with a reported price in the low hundreds of millions of dollars. Bun, founded in 2021 by former Stripe engineer Jared Sumner, has raised at least$26 million across a$7 million Kleiner Perkins seed and a previously undisclosed$19 million Series A led by Colson Ventures.

21:34And its open source runtime now sees more than 7 million monthly downloads despite generating no revenue. Anthropic has already used BUN for over six months to run its own systems and plans to hire seven BUN employees to keep the software open source while integrating it deeply into Cloud Code, its coding agent, to improve performance, stability, and ease the point of deployment for enterprise customers. Cloud Code launched to select customers in February, now offered through Cloud subscriptions and enterprise plans, has reached$1 billion in annualized revenue, and coding-related products represent the majority of Anthropics revenue.

22:15The BUN deal is part of a broader M &A strategy outlined to banks in which Anthropic targets small acquisitions in areas like automated testing and developer tooling. Anthropic generates about 80 % of its revenue from selling API access to its models to app developers, such as Legal AI Startup Harvey and Code Assistant Makers Cursor, and is seeing stronger software tooling as a way to deepen lock-in and grow that API franchise. The prospective deal contrasts with rival OpenAI's more aggressive M &A, which has spent more than$6.4 billion in stock to acquire at least three startups over the past 18 months, underscoring different capital allocation philosophies among top labs.

22:58Anthropic is in early and largely informal talks to pursue an IPO that could come as soon as 2026 and a rank among the largest tech listings ever, even as it continues to raise large private rounds. The company has engaged law firm Wilson Sincini as a veteran of tech IPOs for Google, LinkedIn and Lyft to support internal readiness and has discussed listing options with major investment banks, although it's not selected underwriters. Anthropic spokesperson has stressed that no final decision has been made on timing or whether to go public and that the company already operates internally as if it were public in terms of governance and reporting, which is common at its scale.

23:37investors see a potential Anthropic IPO is a way for the company to seize the initiative from OpenAI, which has been exploring a public listing, but whose CFO recently said that there are no near-term plans despite a$6.6 billion share sale at a$500 billion valuation. Anthropic is simultaneously executing a$50 billion AI infrastructure build-out that includes data centers in Texas and New York and is tripling its international workforce, reflecting capital needs that could ultimately push it towards public markets. The company has also added IPO experienced executives, notably CFO Krishna Rao from Airbnb, who played a key role in Airbnb's 2020 listing and has been expanding its corporate development team to pursue small acquisitions that extend its developer and vertical industry capabilities.

24:24Any anthropic IPO would test investor appetite for large loss making AI companies amid intensifying debate over AI bubbles that would likely be benchmarked against open AI. which now carries an implied private valuation of around$750 billion per softbank. OpenAI has launched a free AI-powered shopping research tool inside ChatGPT that uses a customized version of its GPT-5 mini model to generate personalized buying guides based on user prompts and web reviews with usage described as nearly unlimited for all free and paid users through January. Instead of providing an immediate text answer, the tool presents a quiz-like flow that asks follow-up questions on budget, size, color, and other constraints, and then surfaces 10 to 15 product suggestions while users click more like this or not interested to refine the list, particularly for detail-heavy categories such as electronics, beauty, home, garden, kitchen, appliances, and outdoor items.

25:24Open AI says the model is trained to weigh organic reviews from sources like Reddit more heavily than paid marketing or product page reviews and aggregates feedback across the Internet to generate its guides through. Although it cautions that the tool may still make errors on detailed on details, such as price and availability and urges users to verify information on merchant sites. The tool does not prioritize specific websites and does not share chat data with real retailers. It is also separate from OpenAI's existing in-chat checkout experience, which relies on Stripe and currently supports a limited set of Etsy and Shopify sellers.

26:03This initiative puts OpenAI in direct competition with AI-native shopping search players such as Perplexity. And while it is not yet monetized, the company expects shopping and advertising features to contribute roughly 20 percent of revenue by 2030 as part of a broader push to drive nearly 200 billion dollars in annual revenue by the end of the decade. softbank's 11 stake in open ai acquired for a total of 10.8 billion with a commitment to invest another 22.5 billion provides a window into how the public markets may already value chat gpt developer at roughly 750 billion that's about 50 above the 500 billion valuation used in the company's recent 6.6 billion secondary sale softbank has uh implied that its open ai stake is worth$83 billion, which yields a$756 billion implied OpenAI equity valuation.

26:58OpenAI currently serves more than 800 million ChatGPT weekly active users, of whom about 35 million, or roughly 5%, pay for plus or pro subscriptions at$20 to$20,$200 per month, respectively. and it projects that by 2030, weekly active users will reach 2.6 billion, with 8.5 % of them around 220 million people on paid plans, making ChatGPT one of the largest subscription franchises globally. OpenAI has increased its revenue expectations for ChatGPT and now forecasts around 270 billion in cumulative subscription revenue through 2030, including about 10 billion this year and roughly $87 billion in 2030 alone, as it targets nearly$200 billion in total revenue by decades and across subscriptions, APIs, and new lines such as shopping and advertising.

27:50The company's current non-paying user base drives significant compute costs and has compressed gross margins and internal projections call for cumulative cash burden of about$115 billion between 2025 and 2029, an annual cash burden rising from$17 billion in 26,$35 billion in 27, and$47 billion in 28, necessitating$50 to$80 billion of new capital that likely exceeds what private markets can easily supply. ChatGPT's weekly active user base more than tripled year-on-year in September and remains ahead of Google's Gemini, which has about 650 million monthly active users, where growth has become volatile with 42 % month-on-month jump in January versus December of 2024 and just 13 % month-on-month growth in September versus August.

28:39OpenAI's strategy leans heavily on bottom-up model similar to Zoom and Slack with a free tier that spreads virally among employees and then opens the door to enterprise-wide deals. About 7 million users now access ChatGPT via business plans like ChatGPT Business and ChatGPT Enterprise, which offer enhanced security and data integration for customers such as Canva and PwC. The company is also rolling out age verification by December in response to content safety changes that have slightly reduced usage time and plans to relax some restrictions for verified adults, while its subscription-heavy model contrasts with rival Anthropic, which expects about 80 % of its revenue for API access and only a fraction from Chatbot subscriptions.

29:26OpenAI is entered in a strategic partnership with Thrive Holdings, a portfolio company launched by major OpenAI investor Thrive Capital, under which OpenAI will take an equity stake in Thrive Holdings and embed engineering, research, and product teams directly into its operating companies to drive sector-specific AI development. Thrive Holdings acquires and operates businesses in real economy sectors, such as accounting and IT services, and plans to use OpenAI models and teams to optimize workflow, cut costs, and create new data-driven products across its portfolio with the partnership structure so that OpenAI's stake increases in value as these businesses grow.

30:05OpenAI viewed this as part of a broader strategy of inside-out digital transformation that uses operating companies as labs for model fine-tuning and application development as a mechanism to get compensated for services through equity rather than cash only, only cash fees. Thrive Holding expects to hold assets indefinitely, which aligns with OpenAI's plan to continuously retrain frontier models on company specific data and expert feedback and use those models to create durable enterprise value. OpenAI has agreed to acquire Neptune, an AI tooling startup that builds and monitoring and debugging systems for training large AI models and a deal that will see Neptune wind down its external services and integrate into OpenAI's internal training stack with financial terms undisclosed.

30:55Neptune's software provides fast, precise analytics on complex training workflows through metric dashboards, and OpenAI's chief scientist, Jacob Paktaki, says that tools will deepen visibility into how models learn and enable faster iteration on training regimes. The acquisition follows a string of deals this year in which OpenAI bought software applications Incorporated, Statsig for$1.1 billion, and Johnny Ives' AI device venture, I.O., for more than$6 billion, signaling an aggressive M &A strategy centered on core infrastructure experimentation and user experience. Neptune has raised more than$18 million from investors in its absorption into OpenAI underscores the consolidation pressure on AI tooling startups as hyperscalers and Frontier Labs pull key components in-house.