❯ Crusoe Reportedly Raises About $3 Billion to Expand AI Infrastructure
[Returning Backers] AI infrastructure developer Crusoe raised more than $3 billion at a valuation of roughly $30 billion, co-led by Atreides Management and Valor Equity Partners, Bloomberg reported. It secures power, builds data centers and operates a GPU cloud. Mubadala Capital participated. Crunchbase classifies the financing as Series F and puts total funding at nearly $7.2 billion.
[From Mining to AI] Crusoe’s transition followed a clear asset trajectory. It initially used stranded natural gas to power cryptocurrency mining, then redirected its energy-development experience toward AI. The company says it sold its mining business in 2025, concentrating on data centers and its cloud platform. That October, it announced the initial close of an anticipated $1.375 billion Series E at an expected valuation above $10 billion. The latest reported valuation of roughly $30 billion brings it to around three times that earlier $10 billion threshold. This expansion builds on a business that has already changed direction; treating Crusoe as a cryptocurrency miner would miss what it now sells.
[Power and Cloud Together] Its capabilities have two sides. On engineering, the company says the first phase of its Abilene, Texas, campus went from groundbreaking to operation in less than 12 months. The full campus has a planned capacity of 1.2 gigawatts; that should not be confused with capacity already operational. On software, Crusoe Cloud offers GPU compute and managed services, while the acquisition of Atero strengthened GPU memory management and optimization. Coordinating power, facilities and cloud services within one company can reduce handoffs between builders and operators, but increases funding and management requirements. Customers ultimately buy reliable compute, rather than the capacity figures in a campus plan.
[Turning Speed Into Cash] Funding provides the means to start construction. Power connections, equipment delivery and customer acceptance determine when revenue can begin. A premium for integrated delivery is a bet that complex projects will become paying assets more quickly. The same integration, however, increases exposure to schedules and tied-up capital: a delay at one stage can hold back cash collection across the chain. Assessing this valuation requires looking beyond fundraising or contracted capacity to operations and cash receipts together. Construction capability wins orders; cash recovery determines how long this asset-intensive business can keep expanding.
▪ SIGNALPower access and construction speed can win orders. Sustaining a delivery premium requires converting commissioning schedules into cash receipts.
❯ Fluidstack Reportedly Raises $1.5 Billion for Compute Infrastructure
[A Quant Fund Leads] Compute infrastructure provider Fluidstack closed a $1.5 billion financing led by quantitative trading firm Jane Street at a valuation of more than $18 billion, Forbes reported. It deploys and operates large computing clusters and custom data centers for AI labs. Crunchbase classifies the deal as private equity and puts total funding above $2.6 billion. The latest transaction remains reported information.
[Separate Closing From Disclosure] The previous round needs particular care. An announcement published on the company’s website on July 20 states that its $830 million Series A closed in January at a $7.5 billion valuation, led by Situational Awareness. The publication date should not be presented as the closing date, nor should the $750 million figure in some reporting override the company’s disclosure. Comparing the official earlier valuation with the latest reported figure puts the new valuation at more than 2.4 times the old one. Repeated fundraising within the same year needs to be understood alongside its large construction commitments, rather than attributed entirely to market enthusiasm.
[An Anchor Construction Role] Validation comes from the customer side. In November 2025, Anthropic announced a $50 billion infrastructure plan, naming Fluidstack as its partner for custom data centers in New York and Texas scheduled to come online during 2026. That is the customer’s overall investment plan, not financing received by Fluidstack. Forbes also describes a model centered on building and operating facilities that host customers’ chips, differing from buying GPUs and renting them out. The valuable capability is rapid deployment to customer requirements: matching equipment, power and operations to model workloads so that labs can turn compute demand into usable facilities.
[Equity Is Only One Ledger] Holding fewer chips directly can reduce some hardware-refresh exposure. It does not automatically remove construction, financing or customer-concentration risks. Equity funding, project borrowing and customer purchase commitments solve different problems and cannot be added together as revenue. This round raises questions about who bears construction risk, how customer payments cover fixed expenses and whether contracts support subsequent expansion. Specialized operators retain bargaining power while labs place a premium on delivery speed. If construction slows, valuation will return to contract quality and collections, rather than the size of the development pipeline.
▪ SIGNALBuilding for major AI labs can expand financing capacity. Customer commitments, project funding and operating revenue must still be counted separately.
❯ Gimlet Labs Raises $300 Million to Expand Its Multi-Silicon Inference Cloud
[Strategic Investors Join] Gimlet Labs announced a $300 million Series B led by a16z at a $3 billion valuation. It breaks model inference into tasks that run cooperatively on different types of chips. Total funding is $392 million, with Sapphire Ventures, Menlo Ventures, Arm and Microsoft’s M12 among the participants.
[A Larger Round Within Months] On March 23, Gimlet announced an $80 million Series A led by Menlo Ventures. By early September, its new round was 3.75 times that size. At the Series A, the company said its customer count had tripled in the five months since its public launch, including a frontier model lab and a hyperscaler, neither named. The rapid return to fundraising comes as agents make sequential model and tool calls, accumulating delays at every step. Customers consequently demand faster responses from inference services. The previous round supported demand validation; the new one must expand delivery at scale.
[Different Chips, Different Jobs] The company’s product description goes beyond “inference acceleration software.” Its product decomposes model tasks and schedules them across GPUs, CPUs and other accelerators according to customer workload requirements and available hardware. Reading context and generating an answer place different demands on computation and memory, favoring different chips. Gimlet builds cross-chip orchestration software and needs the corresponding data-center connectivity, making it more than a lightweight software tool. Its September announcement says it added billions of dollars in contracted revenue since March and accumulated gigawatts of data-center pipeline. Those are contracts and planned projects, not recognized revenue and live capacity.
[Efficiency After Complexity] Assigning tasks to better-suited chips could allow more requests to be processed within the same power constraint, giving an inference cloud an architectural source of profit. Mixed hardware also adds networking, integration, operating and utilization costs. Investors should be buying a whole-system cost advantage, rather than a faster result on one stage of a benchmark. Arm and M12 bring industry connections, but customer bills provide the ultimate test: for the same model and service quality, is the cost per useful task lower, and can that advantage survive the next generation of chips?
▪ SIGNALInference-cloud competition is moving deeper into chip specialization. Performance gains become sustainable margins only after paying for system complexity.
❯ Upwind Reportedly Raises $300 Million for Cloud Runtime Security
[Familiar Investors Return] Cloud-security company Upwind raised about $300 million at a valuation of roughly $3.8 billion, co-led by Bessemer Venture Partners and TCV, CTech reported. It combines cloud scans with runtime data to identify exposed vulnerabilities and attacks in progress. Salesforce Ventures, Greylock, Craft Ventures and Cyberstarts were among the participants.
[A Year of Fundraising] Upwind announced a $250 million Series B in January and said in March that Salesforce Ventures was joining that round. This week brought reports of another roughly $300 million financing, classified as Series C by SiliconANGLE. Comparing CTech’s roughly $1.5 billion valuation at the start of the year with the latest roughly $3.8 billion figure gives an increase to about 2.5 times the earlier level; the $1.8 billion starting figure in some reporting is not used here. Crunchbase lists total funding of $730 million. The fundraising sequence is clear, but the reported new round should still be distinguished from the Series B formally announced by the company.
[Risk in Its Operating Context] Upwind starts with how applications are actually running, then prioritizes action. Its website describes live asset inventories, exposure assessment and attack detection, connecting services, identities, network access and data flows. This lets security teams prioritize issues that are both active and reachable instead of relying solely on static severity ratings. The team also has cloud-platform experience: according to the company, its founders previously built cloud-optimization provider Spot.io, which NetApp acquired. The differentiation rests on data from live environments and engineering experience. Whether it consistently reduces the burden on security teams must be demonstrated in customer use.
[Earning a Lasting Budget] Such a platform earns a lasting budget by helping understaffed security teams address genuinely urgent issues. Connecting more cloud and AI services expands coverage but makes false positives, missed threats and permission management harder. For investors, remediation efficiency is closer to commercial value than the number of vulnerabilities discovered: do customers shorten investigations, extend deployment and renew? With valuation rising quickly, sales and implementation must keep pace. Otherwise, broader feature coverage merely adds deployment costs without automatically improving retention.
▪ SIGNALA cloud-security platform gains pricing power by helping customers decide what to fix first, rather than putting more alerts on one screen.
❯ HiddenLayer Raises $100 Million to Protect AI Agents
[Purpose-Built AI Protection] AI-security company HiddenLayer announced a $100 million Series B led by Delta-v Capital. It discovers enterprise AI assets, tests attacks and blocks threats while models and agents run. Ten Eleven Ventures, M12, Booz Allen Ventures and Morgan Stanley participated. Crunchbase lists total funding of $156.2 million; the round’s valuation was not disclosed.
[From Models to Agents] Founded in 2022, the company raised a $50 million Series A in September 2023 led by M12 and Moore Strategic Ventures. The focus then was protecting machine-learning models, monitoring inputs and outputs, and identifying adversarial attacks. Three years later, the new financing is twice as large, while the protection extends to agents that write code, call tools and execute sequences of tasks. A bad output can now lead to an actual operation. Enterprises therefore need checks during execution, beyond a test before deployment. The financing supports that product evolution.
[Securing Execution] HiddenLayer’s website brings discovery, AI supply-chain checks, attack simulation and runtime protection into one product suite. This can connect pre-deployment testing with production monitoring and track the same asset through different stages. The new funding will strengthen agent runtime defenses and execution-environment security for autonomous coding agents. In its financing announcement, the company says annual recurring revenue grew more than tenfold over the past year and it added more than 50 platform customers. It did not publish the revenue base alongside that growth figure, so revenue cannot be inferred from the multiple alone. Commercial validation also requires knowing which capabilities new customers use and whether they keep paying.
[Testing a Standalone Category] To command a separate budget, AI security must demonstrate capabilities that existing security products do not adequately provide. Model inputs can change system behavior, while agents connect that behavior to enterprise tools, making task and permission awareness necessary. Yet scanning, testing and runtime protection could also be absorbed into established platforms. HiddenLayer’s opportunity is to deliver hard-to-replace detection while keeping integration and response straightforward. More operating disclosure is needed before assessing valuation. For now, renewals and expansion keeping pace with the product scope would validate the business more convincingly than the AI-security label alone.
▪ SIGNALAs AI moves from answering questions to executing tasks, security products must protect execution to earn independent, recurring budgets.
❯ HiBob Secures $166 Million to Connect Workforce Systems With AI
[An Enterprise Platform Invests] Workforce-software provider HiBob secured $166 million in funding led by Salesforce, with Farallon participating. Its Bob platform helps businesses manage employees, payroll, benefits and workforce planning. The company confirmed the investors, while CTech reported a valuation of roughly $3.2 billion. Crunchbase lists total funding of $740 million; the company did not assign a lettered round.
[Three Years of Accumulation] HiBob raised $150 million in 2023, when it reported more than 3,500 customers. Planned uses included larger enterprise accounts, more modules and wider geographic coverage. The latest announcement puts the customer count above 5,500. Using CTech’s valuation figures for the two rounds, the move from roughly $2.7 billion to roughly $3.2 billion is an increase of about 19%, without the rapid doubling seen in compute infrastructure. Its trajectory is one of expanding business systems and its customer base, then bringing in a strategic investor to find new uses for existing workforce data.
[Making Organizational Context Available] Bob’s existing products cover people, compensation, performance and planning. Those records can answer which team someone belongs to, whom they report to and how staffing budgets change. A continuously maintained business system can offer traceable organizational context more readily than information assembled ad hoc. The company now describes an open-platform direction that brings this context into collaboration, CRM, finance and operations, including integration with Slack. Existing functionality and future aims need separating: HR management is already sold, while broad access to organizational context through agents remains an expansion direction. Salesforce’s involvement supports cross-system cooperation; it does not establish that every integration has shipped.
[Data Needs Boundaries] An AI agent handling work for employees needs to know who can approve a task and who is accountable for its outcome. HiBob’s opportunity is to extend the usefulness of workforce records into everyday business, beyond visits to an HR application. Organizational information also carries privacy and permission constraints, so access requires clear boundaries. Investment returns ultimately depend on additional spending and retention: will customers pay separately for new capabilities, or consider them upgrades that should come with existing subscriptions? Accumulated data has value. Products and contracts must still determine how it earns money.
▪ SIGNALAs enterprise AI enters more workflows, organizational relationships and authorization context become more useful. Additional revenue depends on the new use cases.
❯ Lyte Raises $165 Million to Scale Robotic Perception Production
[Capital for Production] Robotic-perception company Lyte announced a $165 million Series C led by Maverick Silicon at a $1.6 billion post-money valuation. It combines custom chips, sensors and software to help robots determine their position and understand motion around them. Total funding is $272 million; the capital will expand production and deployment.
[Research Reaches Customers] Lyte was founded in 2021 by Alexander Shpunt, Arman Hajati and Yuval Gerson, who previously worked on Apple’s sensing technologies. It emerged from stealth in January with $107 million in aggregate prior funding. September’s Series C exceeds that entire earlier total. The company says it has entered production and is shipping to robotics customers in inspection, logistics and manufacturing. Spending is moving beyond extended research toward manufacturing and customer deployment. Fidelity, which led the Series B, returned alongside Atreides Management, Key1 Capital, Ora Global and other investors.
[Aligning Perception First] The company’s Lyte Vision product combines 4D sensing, color images and inertial measurements. The 4D sensing includes distance and velocity, allowing robots to understand where objects are and how they move. Lyte Galaxy organizes sensing, compute, algorithms and software into a unified perception platform. The approach coordinates hardware and data first, reducing the work robot makers must do to assemble and synchronize disparate sensors. The team’s Apple and PrimeSense experience provides depth-sensing engineering credentials. Production robots operate in different environments from consumer electronics, however, and becoming a dependable supplier still requires prolonged validation at industrial customer sites.
[Beyond One Robot Design] A perception supplier can serve robots with different shapes and tasks, spreading exposure across applications rather than selling a complete robot. A common platform must still handle differences in lighting, movement, occlusion and vibration; a single demonstration cannot settle those questions. The most meaningful evidence following this round will be volume deliveries and repeat orders, along with manufacturing yields and unit costs that support gross margins. If robot makers outsource perception, Lyte can participate in growth across several product lines. If every project requires extensive customization, economies of scale will take longer to emerge.
▪ SIGNALPerception could become a business spanning multiple robot platforms. The move from prototype to production supplier must be validated through costs and repeat orders.
OUTLOOK
[Funding Concentrates in Delivery] Using rounded amounts from the announcements and reporting, the seven selected deals total approximately $5.531 billion. Crusoe, Fluidstack and Gimlet Labs account for roughly $4.8 billion, or 87%. This is the selected sample, not a global funding census. Funding is more concentrated than the range of themes suggests: security, organizational data and perception are attracting capital, but the largest checks still go to computing infrastructure. Their paths to returns differ. Data centers need commissioning and collections, inference clouds need efficiency, software needs renewals, and perception hardware needs production scale. If delivery succeeds without durable margins, the commercial thesis shared by these financings will be challenged. Cash recovery in each business, rather than valuation rankings, is the next test.