❯ SoftBank is reportedly seeking up to $100 billion from Gulf investors for a fund to buy companies and overhaul them with AI
Son turns to Gulf capitalAccording to Finimize, citing the Financial Times, SoftBank CEO Masayoshi Son is in talks with Gulf investors, including in the UAE, to raise up to $100 billion. SoftBank did not comment, and Reuters said it could not independently verify the report.
Bet big early, then find partnersFinimize sums up Son’s long-running model as placing concentrated bets early and then looking for partners to scale them. SoftBank has mainly invested from its own balance sheet, sometimes with pricey debt. The new fund would manage outside backers’ money, a structure closer to a traditional private equity fund.
Buy companies, then rebuild themBy the Financial Times’ account, the Gulf-backed fund would buy companies and then modernize their operations with AI and other advanced technology. The backdrop is that SoftBank just completed a $30 billion investment tied to OpenAI’s latest funding round; Reuters reported that to help pay for it, the group sold $11.1 billion of high-yield bonds last month, the largest such corporate sale globally.
Less expensive debtIf the Gulf money comes through, SoftBank could rely less on high-interest bonds for its next round of AI-related acquisitions, easing its own refinancing pressure. If the talks fail, the pace of its expansion stays tied to conditions in the bond market. What the backers would be buying is SoftBank’s ability to pick companies and then rebuild them with AI.
▮ SIGNALAI investing has extended from backing model companies to buying traditional businesses and reworking them directly, and the capital required has moved from venture scale to buyout-fund scale.
❯ AI cloud provider Lambda is in talks for a pre-IPO round of up to $4 billion at a $14.5 billion pre-money valuation, after its backlog grew to $50 billion in three months
The last private roundAccording to Yahoo Finance, relaying a Wall Street Journal report, AI cloud infrastructure company Lambda is negotiating a pre-IPO round of up to $4 billion at a $14.5 billion pre-money valuation, led by Blackstone and Coatue. Lambda spokespeople declined to comment on the fundraising details.
Buying GPUs to rent them outLambda belongs to the group known as neoclouds: companies that acquire high-performance GPUs and lease them to enterprises training and running large models. It relies exclusively on Nvidia chips and server architecture, and Nvidia took part in its $480 million Series D in February 2025. Last November, TWG Capital led a $1.5 billion round that valued it at $5.9 billion post-money.
One contract carries most of the backlogAn investor document reviewed by the Journal shows Lambda’s backlog growing from $15 billion in June to $50 billion in September, of which $35 billion comes from a cloud services agreement with Anthropic, using Nvidia hardware hosted at a facility in Texas. Management is targeting a 2027 listing, contingent on execution and market conditions.
A new team for the listingIn May, co-founder Stephen Balaban moved to chief technology officer and Michel Combes, a former chief executive of Sprint and Alcatel-Lucent, became CEO; the company also added a former AT&T chief executive to its board. With orders so concentrated in one customer, new investors have to judge whether that contract converts to revenue on schedule and whether Lambda can diversify its customer base before going public.
▮ SIGNALCompute-leasing companies are increasingly valued on the long-term contracts they hold, and with those contracts concentrated in a few model companies, their fates are tied together.
❯ Nvidia-backed Firmus withdraws its roughly $5 billion Australian IPO and turns to private markets
The board pulls the offerAustralian AI data center operator Firmus has withdrawn its planned listing. In a statement to CNBC, the company said its board determined that the proposed offering terms did not adequately reflect the strength of its business and long-term growth outlook and that proceeding was not in the best interests of the company and its shareholders, also citing market volatility. Firmus said it will pursue capital from private markets and consider other public and private options.
It would have been Australia's second-largestFirmus had reportedly planned to price shares at A$11 and raise about $5 billion, implying a valuation of about $30.6 billion. By size of new shares sold, it would have been the second-largest in Australian history.
Nearly triple in two monthsIn August, Firmus announced a $2 billion round backed by Nvidia, Coatue, Blackstone and Jane Street, bringing equity raised over the preceding year to more than $3 billion and its valuation to over $10.5 billion. The IPO price implied nearly three times that. Last month the company also announced agreements with Meta to provide GPU capacity at its AI data centers in Southeast Asia.
Same investors, two outcomesBlackstone and Coatue are leading Lambda’s private round this week while watching Firmus, which they also backed, pull out of the public market. Private investors are willing to price compute companies on long-term contracts; public buyers this time did not accept the price Firmus proposed. Peers preparing to list next year will use this withdrawal as a reference when setting their own terms.
▮ SIGNALCompute companies have grown more expensive with every private round, and only at the door of the public market did they first meet buyers unwilling to pay the asking price.
❯ Flex brings in a $2 billion preferred investment for Axiom ahead of its spin-off, setting enterprise value at $37.5 billion
A price before the separationManufacturing services provider Flex has secured a $2 billion convertible preferred investment in its Axiom Solutions International unit. According to Converge Digest, the buyers are funds affiliated with General Catalyst and Koch Equity Development, along with co-investors, and the deal establishes an initial enterprise value of $37.5 billion for Axiom.
A power chain from grid to chipAxiom is the business Flex has assembled around high-density AI infrastructure, combining embedded and critical power, electrical infrastructure, power conversion, thermal management and compute integration. Flex formally named the business in September and plans to spin it off as an independent public company in the first quarter of 2027, subject to required approvals.
The money goes to EPC PowerThe financing provides equity capital for Axiom’s pending acquisition of EPC Power, a $4.4 billion deal Flex announced in September, with committed term loans arranged for the remainder. EPC Power moves Axiom’s portfolio upstream into megawatt-scale power conversion, energy storage and 800-volt DC architectures. The preferred shares carry a 10% annual cash dividend before the separation.
A benchmark for power hardwareAxiom had previously disclosed its operating scale but had no valuation set by outside investors. The $37.5 billion figure gives its spin-off a reference point that public investors will later compare against its trading price. Neither the preferred investment nor the EPC Power acquisition has closed.
▮ SIGNALThe hardware that powers and cools data centers is being priced and listed on its own, as capital values AI infrastructure layer by layer beyond the chip.
❯ Spend management company Ramp reportedly raises about $1.85 billion at a $60 billion pre-money valuation, up $16 billion in four months
Nearly double what was first discussedAccording to AI Weekly, relaying an October 9 Bloomberg report, Ramp closed about $1.85 billion led by Dragoneer and Thrive Capital, with Founders Fund also writing a large check, at a $60 billion pre-money valuation. When Bloomberg reported on it in early September, the company was in talks for about $1 billion.
Managing how companies spendRamp made its name managing corporate expenses. As PYMNTS recounts, it launched Ramp Stack, an AI system for accounting firms, in June, and in August launched Router, which decides which AI model handles each task for a company; Ramp said the tool cut its own AI costs by about 30%.
Three prices in under a yearLast November, Ramp was valued at $32 billion when it closed a $300 million financing and an employee tender offer; in June it raised a $750 million Series F at $44 billion; now it is at $60 billion. The company said in June that total payment volume grew about 170% year over year in March, its highest growth rate in three years.
Finance software starts managing the AI billRamp’s customers are corporate finance departments. As AI spending becomes a large line item of its own, whoever can see and control that money holds a new entry point, and that is what Router does. Traditional expense and spend-control software that does not follow will lose its place in the fastest-growing category of spending.
▮ SIGNALA spend-control company is telling its growth story around saving customers money on AI, which shows enterprise AI spending is now large enough to need dedicated tools to manage it.
❯ TypeSafe AI, maker of the Jev model, raises an $870 million Series A at a $7.5 billion valuation less than a month after launch
An $870 million Series ATypeSafe AI announced on October 9 an $870 million Series A at a $7.5 billion valuation. Andreessen Horowitz led the round, with Sequoia Capital, existing investor DCVC and a group of angel investors participating, and a16z’s Martin Casado is joining the board.
A model that outputs no textJev is based on a transformer architecture but is not a large language model: it produces probabilities, not text. As SiliconANGLE describes it, Jev supports just three types of requests: answering yes or no, picking an item from a list, or generating a score, each with a measure of its confidence. After calling an LLM, enterprise software usually has to condense a block of text into a structured format before using it; Jev skips that step.
Launched three weeks agoAccording to TechCrunch, Jev was released on September 15. The company was founded in 2024, and co-founder Diogo Almeida was previously a researcher at OpenAI. TypeSafe says Jev processes a request in under 700 milliseconds and that a third of the Fortune 500 already use it; these figures are the company’s own.
General-purpose LLMs lose some callsThe company’s announcement says it will ship more “machine-native” models and add the enterprise features customers have asked for. Many automation steps inside companies need a decision, not a paragraph. If those steps move to specialized small models, general-purpose LLMs billed by the token will lose part of their enterprise call volume.
▮ SIGNALA model that cannot talk reached a $7.5 billion valuation in three weeks; investors are betting that in enterprise automation, deciding is more common than generating and more sensitive to speed and cost.
❯ AI drug discovery company Isomorphic Labs is reportedly in talks for new funding at a $40 billion to $50 billion valuation
A valuation before a deal sizeAccording to SiliconANGLE, citing a Bloomberg report on October 8, Alphabet spinoff Isomorphic Labs is in talks for a new funding round at a valuation of $40 billion to $50 billion. The talks are at an early stage, and neither the size of the round nor the investors have been disclosed.
Finding where a drug can landIsomorphic’s flagship product is IsoDDE, an AI platform that speeds up the manual research steps in developing a therapy. A drug usually has to attach to a specific point on a disease-causing cell, known as a binding pocket, and finding one is slow and difficult. The company said in February that IsoDDE was more than 50% better at finding binding pockets than the closest competing tool.
Revenue from drugmaker partnershipsThe company earns money through research partnerships with drugmakers. In 2024 it signed deals worth nearly $3 billion with Eli Lilly and Novartis, and this January it launched a similar collaboration with Johnson & Johnson; all three focus on small molecules. Its last raise, a $2.1 billion round announced in March, included Alphabet, two of its venture funds and several outside backers.
Multiples higher in little over six monthsOnly a little over six months separate the $2.1 billion round in March from the tens of billions now being discussed. Investors are betting that platforms like IsoDDE can sharply cut the time needed to screen drug candidates. Lilly, Novartis and J&J have already signed on, and the more capital favors the platform, the harder it becomes for later drugmakers to get the same terms.
▮ SIGNALAI drug discovery is now valued on the scale of early rounds at leading model companies, but its proof arrives on the timetable of clinical trials, which will test investors’ patience far longer than software does.
❯ Manus closes a round of more than $500 million at a reported $4 billion post-money valuation, twice its previous price
The missing valuation arrivesManus parent Butterfly Effect announced on October 8 a round of more than $500 million led by Boyu Capital and IDG Capital, with existing shareholders Tencent, HongShan (Sequoia China) and ZhenFund participating again, without disclosing a valuation. LatePost reported the same day that Manus now has a post-money valuation of $4 billion.
An agent that finishes the jobManus launched in March 2025 and went viral on a demo of its AI agent. As TechCrunch describes it, the company offers a chatbot and tools for building apps and websites in natural language, and can also produce designs, presentations and video. When Meta announced its acquisition in December 2025, Manus was reportedly bringing in more than $100 million in annual recurring revenue.
The first money since going independentMeta’s purchase price was about $2 billion; the deal was not completed, the founding team and existing shareholders bought the shares back from Meta, and the company said on September 1 that it had resumed independent operations. According to SiliconStar, this is the largest single round to date in China’s AI application sector.
More cash, more runwayLatePost says Manus is the most highly valued AI application company with a China background, with the most cumulative funding and the largest cash reserves. SiliconStar notes that it released Manus 2.0 and a personal agent app called Cue on September 28, is developing a product for the Chinese market, and has opened 17 positions. Leading on cash reserves means it can outlast comparable application companies on hiring and compute.
▮ SIGNALAn application company’s valuation doubled after an acquisition that did not close; investors no longer need a big-tech buyer to anchor the price of the agent layer.
❯ Quantum computing company Oratomic raises a $475 million Series B, lifting its valuation from $1.5 billion to $5.5 billion in three months
Six firms co-leadCalifornia quantum computing company Oratomic announced a $475 million Series B on October 8. According to SiliconANGLE, ARCH, Spark Capital, Khosla Ventures, Index Ventures, General Catalyst and Bezos Expeditions co-led the round, bringing total funding to $775 million and lifting the valuation from $1.5 billion to $5.5 billion.
Qubits from neutral atomsFounded in 2025, Oratomic builds qubits, the basic computing units of a quantum computer, from reconfigurable arrays of neutral atoms. Qubits are extremely fragile: slight disturbances in heat, electromagnetism or vibration can make them lose information, so error correction is needed. The industry generally assumes about 1 million qubits for commercially meaningful fault tolerance; Oratomic believes it needs only 10,000.
The last round was three months agoIts Series A was $300 million, just three months before this one. CEO Dolev Bluvstein told the Wall Street Journal the company has no idea what the business models or exact applications will be, only that it must first build a working computer that is cheap and manufacturable.
Other approaches will be measured against itThe same day, UK-based Universal Quantum, which uses trapped ions, announced a Series A of more than $100 million. SiliconANGLE cited McKinsey data showing the quantum computing industry generated just over $1 billion in revenue worldwide in 2025. With a one-year-old company valued in the billions, investors will compare technical approaches on metrics such as qubits required, and the bar for raising money rises for other companies too.
▮ SIGNALQuantum fundraising is starting to move at the pace of AI model companies, but without revenue to verify it, the valuation rests on an engineering hypothesis that has yet to be proven.
❯ Data center rack maker Oxide Computer raises a $445 million Series D, with demand still outrunning a 20x capacity expansion
Returning backer Eclipse leadsData center hardware company Oxide Computer announced a $445 million Series D on October 9. According to SiliconANGLE, returning backer Eclipse led, joined by AMD Ventures, Jane Street and others. According to Techmeme, relaying Axios, the round values the company at $6 billion and brings total funding to about $835 million.
A rack that arrives assembledOxide sells a data center rack that combines computing, storage and network equipment in a pre-integrated package. Each rack has 24 hardware modules called sleds, each hosting an AMD processor with up to 144 cores. Customers can split the hardware in a rack into multiple virtual networks; a software company, for example, could give each engineering team its own environment.
The previous round was in FebruaryThis February, Oxide closed a $200 million round led by USIT. CEO Steve Tuck said on-premises cloud projects are accelerating, that the company scaled manufacturing capacity 20x over the past 12 months, and that demand still exceeds supply; the money will go into materials and manufacturing scale.
Bringing the cloud back in-houseOxide’s customers are companies that want a cloud experience in their own data centers. In August it disclosed plans for a computing sled that can host a graphics card and hinted it would partner with AMD. That would give companies one more way to run AI workloads on premises without assembling servers themselves, and leave public cloud providers facing a share of compute demand that stays in customers’ own hands.
▮ SIGNALWith compute tight and data required to stay on site, the business of buying the cloud and taking it home is growing again, and manufacturing capacity is now its biggest constraint.
❯ SignSplit exits stealth with $400 million in seed funding at a $1 billion valuation to license AI training data and share the proceeds
$400 million on day oneData licensing company SignSplit emerged from stealth on October 5 with $400 million in funding at a $1 billion valuation. According to SiliconANGLE, the backer is W Group, a group of 11 financial technology firms focused on blockchain.
Sign the data, split the valueSignSplit’s concept is “signed data”: human data contributions with established provenance and clear consent to be used under defined terms. Individuals can digitally sign their datasets, likeness, voice or bodies of work, setting the conditions under which AI models may use them; or they can contribute to “data pools” that aggregate data for AI developers and are compensated based on how often it is used.
Quietly founded in 2024The company was established in 2024 and is led by co-founder and CEO Alessandro Monterosso. Its premise is that the next generation of AI and intelligent robotics needs high-quality real-world human data, while most existing models rely on internet-scraped content of opaque origin, so developers will have to pay for better data. W Group said it will devote additional capital and resources to a global rollout.
Companies can build their own poolsSignSplit also targets businesses: a company can create its own data pool, define the real-world data or skills it needs, and receive contributions directly from individuals who agree to its terms; organizations can also use its verification layer to check whether a dataset or likeness is signed. For AI developers, training data gains a source with a licensing record, at the cost of paying for content that used to be scraped for free.
▮ SIGNALThe provenance and licensing of training data is becoming an infrastructure business, and whoever can issue a verifiable record of origin stands to collect a toll in the paid-data market.
❯ Engineering simulation company Vinci raises a $250 million Series B at a $1.5 billion valuation, starting with chip thermal simulation
Three firms co-leadVinci announced on October 6 that it closed a $250 million Series B at a $1.5 billion valuation. According to SiliconANGLE, Advent, Temasek and Xora jointly led, joined by AMD Ventures and others. The company is based in Palo Alto, California.
A simulation after every design changeHardware engineers use simulations to find issues in their designs; a car designer might simulate fog to evaluate its effect on a vehicle sensor. Setting up simulations has historically required highly specialized skills and a lot of time, so many teams use them only intermittently. Vinci’s platform can run a simulation after every major change to a blueprint, surfacing issues the next day instead of two weeks later.
Minutes for what took daysVinci is starting with chip projects, helping engineers simulate the thermal properties of processors, memory modules and interconnects. The platform is powered by a physics-focused AI model: upload a chip’s CAD drawing and related data, and the AI turns it into a simulation and performs meshing, the most time-consuming step, automatically. The company says simulations with hundreds of millions of degrees of freedom finish in a few minutes; such workloads usually run for hours or days.
Next come vehicles and satellitesCEO Hardik Kabaria said the company is starting by letting engineers understand physical behavior while they are still designing. Vinci plans to extend its simulation capabilities beyond chips to vehicles, aircraft and satellites. As simulation shifts from occasional to routine, chip teams find problems earlier and product development times shorten.
▮ SIGNALAI’s first foothold in hardware design is simulation, not generating blueprints, because the bottleneck there is slow computation, and computing fast is what models do best.
❯ Atomic Machines exits six years of stealth with $250 million raised to date to build micro-machines directly from code
$250 million over six yearsAtomic Machines, based in Emeryville, California, emerged on October 7 from six years in stealth and said it has raised $250 million to date. According to FourWeekMBA’s account of the company’s release, investors include OneIM, Gigafund, XTX Ventures, Valor Equity Partners and the Regents of the University of California. The company gave no valuation or breakdown by round.
Compiling code into machinesThe company introduced the Matter Compiler, which it calls an AI-native, all-digital manufacturing system that builds working machines directly from code. It is aimed at micro-machines: motors and gears the size of a grain of sand, robots small enough to work inside the body, and cooling built directly into AI chips. The release says the system builds moving parts with features down to single-digit microns with no hard tooling.
The first product goes to data centersFounder Jeff Holden was Amazon’s tenth engineer and later Uber’s first chief product officer. The company’s first product, PrimeSwitch, is an electromechanical relay aimed at AI data centers moving to 800-volt DC distribution. The company says it opens in 50 microseconds and will be shown publicly for the first time at the OCP Global Summit on October 12–15. These performance figures are the company’s own.
One line, many devicesAs Dealroom describes it, the industry has long worked on “one product, one process,” with each new class of micro-device taking years of custom development. If one system can really produce different devices by changing code, the time and cost of developing a new device both fall. What it has to prove first is that PrimeSwitch can be produced at volume to its stated specifications and sold into data centers.
▮ SIGNALInvestors are carrying the software logic of shipping a new product by changing code into physical manufacturing, and its first customer is still the best-funded buyer around: the AI data center.
❯ Model evaluation platform Arena raises a $200 million Series B at a $3.1 billion valuation, with annualized revenue above $100 million
Valuation nearly doubles in ten monthsAI model evaluation platform Arena announced a $200 million Series B on October 8 at a $3.1 billion valuation. Lightspeed and Khosla Ventures co-led, with Salesforce Ventures, Dell Technologies Capital and others participating alongside existing investors including a16z and Felicis.
Users vote, models get rankedArena began in 2023 as a research project at UC Berkeley. As TechCrunch describes it, ordinary users enter prompts or request coding projects for free and judge which model did better, and those votes feed the leaderboard. In September last year it introduced a paid product, AI Evaluations, selling model labs and enterprises detailed performance analytics based on community feedback.
Revenue more than tripledThis January, Arena closed a $150 million Series A at a $1.7 billion post-money valuation, when annualized revenue was $30 million. The company now says annualized revenue has passed $100 million. Alongside the funding it released the Alignment Index, which measures whether models take unauthorized actions, misattribute statements or falsely claim to have completed tasks.
The referee's customers are the playersArena argues that static benchmarks break down once models recognize they are being tested. Enterprises buying models therefore need evaluations drawn from real use, and model labs are willing to pay to learn their own weaknesses. Arena’s own difficulty is staying neutral: its paying customers are the same labs it ranks.
▮ SIGNALAn evaluation company more than tripled revenue in a year; vendor-published scores are no longer enough to buy on, and refereeing has become a business of its own.
❯ Fusion company Type One Energy raises a $200 million Series B to advance its 400 MW plant project in Tennessee
Climate funds and industrial capital arrive togetherFusion company Type One Energy announced on October 6 the completion of a $200 million Series B co-led by Breakthrough Energy Ventures and Clutterbuck Capital, with new investors including Lowercarbon Capital, Siemens Energy Ventures and SiteGround Capital.
The stellarator routeType One builds stellarators, a type of fusion device. The company believes the principal challenges to commercializing stellarator technology are now engineering and industrial, not questions of plasma physics. Its Project Infinity has two parts: the Infinity One engineering prototype, and Infinity Two, a 400 MWe commercial fusion power plant planned for the Tennessee Valley Authority’s Bull Run site.
Not building everything in-houseThe company uses a partner-based business model: drawing on the expertise, resources and existing production infrastructure of established energy companies instead of recreating them internally, to reduce development costs and risks. Siemens Energy Ventures’ participation reflects that structure. Before the round, the project received its initial fusion operating license from the State of Tennessee.
The money goes to design firstCEO Christofer Mowry said the financing lets the company stay focused on advancing its stellarator technology and Project Infinity design activities while working with industrial partners to deliver the first commercial plant. The plant is sited at TVA’s Bull Run; under the company’s model, manufacturing, engineering and operational capabilities come from partners, and the $200 million mainly supports technology development and design.
▮ SIGNALFusion companies are starting to present industrial partners and licenses as selling points when raising money, as the threshold for this business moves from whether the physics works to whether the engineering can be built.
❯ Specialty insurer Ledgebrook raises $200 million co-led by Allianz’s investment arm, using AI to cut quotes from weeks to hours
Allianz invests and reinsuresBoston-based specialty insurer Ledgebrook announced on October 7 that it raised $200 million in equity financing co-led by Allianz X, the strategic investment arm of the Allianz Group, and Rockefeller Capital Management. Separately, Allianz Re agreed to enter a multi-year reinsurance agreement with it.
Pricing hard-to-place risksLedgebrook writes general liability, professional liability and specialty cover for mid-sized businesses, handling hard-to-place risks that standard carriers will not take, and distributes exclusively through wholesale brokers. Its proprietary AI underwriting platform, Blackbird, reads submissions, classifies risk and calculates a technical price, often quoting in hours instead of weeks, with experienced underwriters making the final decision.
Nearly $1 billion in premium in three yearsThe company started writing business in 2023 and is on track to surpass $1 billion in cumulative premium. An executive said admitted carriers have tightened their appetite and pushed risks once routinely placed with them into this market, driving multi-year growth. The funds will be used to scale its technology and extend Blackbird.
AI goes to the hardest policies firstAllianz X CEO Nazim Cetin said the industry is applying AI to its simplest risks first because that is where automation comes easily, while the larger prize sits in complex, hard-to-place business, where underwriting judgment is scarce and slow decisions cost the most. For brokers, quotes arriving in hours instead of weeks mean difficult placements find a carrier faster.
▮ SIGNALAn insurance giant is tying itself to an AI underwriter through both investment and reinsurance, which suggests what it most wants to buy is underwriting speed, not just a financial return.
OUTLOOK
Private markets set the price, public markets push backTaken together, this week’s deals show a split in the compute trade. Lambda negotiated a $14.5 billion pre-money valuation on a $50 billion backlog, and Axiom received a $37.5 billion enterprise value ahead of its spin-off; both prices came from private investors. Firmus took the same story to the public market and did not get the price it wanted. The logic in applications and models is more direct: Ramp and Arena point to payment volume and annualized revenue, and TypeSafe to enterprise adoption. The money going to deep-tech companies such as Oratomic, Atomic Machines and Type One Energy still corresponds to engineering goals that have yet to be proven. Over the coming months, how compute companies preparing to list are priced will test the valuations private markets have set in this cycle.