AI Market Intelligence Brief: Week of Feb 17-24, 2026

Data by AI Market Watch

AI Market Intelligence Brief: Week of Feb 17, 2026 - Feb 24, 2026

Executive Summary

The AI market has permanently bifurcated into a "barbell" structure, characterized by capital-intensive foundational infrastructure at one end and hyper-specialized, agentic vertical applications at the other. This week's staggering $66.8B in deployed capital across our top 10 tracked companies perfectly illustrates this reality. OpenAI's monumental $63.9B Series F round sucks the oxygen out of the generic model layer, signaling that the foundational wars are now the exclusive domain of sovereign-level capital and hyperscaler balance sheets. To compete at the base layer in 2026 requires tens of billions; anything less is structurally unviable.

The dominant theme of the week is the collision of artificial intelligence with physical reality. We are witnessing a rapid transition from digital-only generative AI to "Physical AI" and spatial intelligence. World Labs securing a $1B strategic round for 3D Large World Models, alongside Freeform's $67M Series B for AI-native metal manufacturing, proves that the next frontier of value creation lies in software-defined physical production and spatial reasoning. The market is aggressively pricing in the reality that bits must now govern atoms, moving beyond 2D text and image generation into high-fidelity, persistent 3D environments and autonomous factory floors.

Furthermore, we are witnessing the absolute death of the generic LLM wrapper. Capital is flowing exclusively into agentic workflows and infrastructure that boast insurmountable switching costs. Companies like Jump, which just raised an $80M Series B to embed an AI operating system directly into the CRM and compliance workflows of 27,000 financial advisors, demonstrate that workflow integration is the ultimate defensive moat. The market rewards platforms that act as verifiable systems of record rather than mere systems of engagement.

For our LPs and portfolio strategy, the implication is a strict mandate: we must execute the barbell strategy. We will concentrate our deal flow either on the infrastructure bottlenecks that enable the mega-caps—such as Fluidstack's distributed GPU aggregation—or on deeply embedded vertical operating systems with proprietary data flywheels. We will immediately pass on any application-layer AI that lacks formal mathematical verification, physical infrastructure integration, or deep workflow embedding. Capital is too expensive to fund generic AI experimentation.

Market Context

The macro environment in early 2026 is defined by a severe bottleneck shift: the primary constraint on AI scaling is no longer silicon supply, but power generation and grid interconnection. With interconnection queues stretching 7 to 10 years and new 2026 state regulations enforcing strict "growth pays for growth" tariffs on data centers, energy availability is dictating market winners. This macro reality directly explains the massive signal strength behind Emerald AI. By orchestrating GPU workloads to reduce power consumption by 25% during grid stress events, Emerald provides a software-only solution to a hardware-scale crisis. They unlock up to 100 GW of grid capacity on existing infrastructure without waiting for decade-long transmission upgrades, solving the exact pain point hyperscalers are currently facing.

Simultaneously, we are tracking a massive wave of vertical integration at the foundational compute layer. OpenAI's confirmed 10-gigawatt custom AI accelerator deal with Broadcom—entering mass production in 2026—signals a definitive pivot away from pure Nvidia dependency toward full-stack hardware control. This hyperscaler drive for infrastructure sovereignty perfectly contextualizes the momentum behind Fluidstack. As traditional cloud providers struggle with multi-month wait times for compute, Fluidstack's distributed bare-metal OS allows them to deploy multi-thousand GPU clusters in 48 hours, capturing sovereign-level deals like their €10B commitment for a 1GW AI supercomputer in France.

Finally, the explosion of AI agents, MCPs, and browser-based developer tools has created a massive security blind spot that traditional EDR platforms cannot see. The cybersecurity industry is shifting from "alerting" to "acting." This macro trend is driving the rapid capitalization of companies like Cogent Security, which uses agentic AI to autonomously remediate vulnerabilities, and Koi, which secures the non-binary software supply chain. The market is demanding security solutions that remove manual triage work entirely, as enterprise security teams can no longer scale headcount linearly with the exponential rise in AI-generated threat vectors.

Funding Spotlight

This week's funding data reveals a clear flight to quality and scale, with deal size distribution highlighting a rapidly maturing market. The sheer magnitude of OpenAI's $63.9B Series F—backed by Microsoft, SoftBank, and Thrive Capital—demonstrates that investors view frontier AGI not as a speculative venture, but as utility-scale infrastructure. However, beneath this mega-round, capital is disproportionately flowing into categories that solve specific, high-value enterprise bottlenecks: Computer Vision ($1.2B to World Labs) and specialized AI Infrastructure ($450M Series B to Fluidstack). Investors are signaling that infrastructure still commands premium capital, provided it delivers a distinct speed-to-market or physical capability advantage over incumbent hyperscalers.

The patterns investors are signaling point heavily toward "verifiability" and "defensibility." In the Generative AI category, Black Forest Labs' $300M Series B (bringing total funding to $481M) is a masterclass in market positioning. Investors like a16z and Salesforce Ventures are rewarding BFL's dual-license open-weights model because it captures both developer mindshare (400M+ downloads) and enterprise budgets ($140M Meta contract). This signals that in the model layer, open-weights combined with superior technical architecture is the only viable commercial moat against closed-ecosystem incumbents like Midjourney and Google.

At the Series A and B stages, we are seeing aggressive capital deployment into Cybersecurity AI and Fintech AI, where deal sizes reflect high conviction in agentic workflows. Jump's $105M total funding illustrates the premium placed on deep workflow integration in wealth management. Similarly, Cogent Security's $42M Series A signals that the cybersecurity market is moving past anomaly detection into autonomous remediation. Deal size distribution confirms that the Seed stage is now reserved for massive TAM zero-to-one plays—such as Kontigo's $20M Seed round to build a USDC-smart neobank for 4.8 billion unbanked globally—while Series A and B rounds require proven unit economics and enterprise-grade compliance.

VC Score Leaders

Our proprietary VC scoring highlights a distinct pattern among the top-tier assets: the highest grades (A+) are awarded to companies exhibiting exceptional founderMarketFit paired with unassailable defensibility. The average VC score across our top evaluated companies sits at a stellar 91/100. Black Forest Labs leads the pack with a 94/100 overall score and a near-perfect 99/100 in aiCapability. Their moat is built on the fact that their founding team literally invented the Latent Diffusion Model architecture. This technical pedigree translates directly into a structural advantage in text rendering and anatomical accuracy that generic wrappers cannot replicate.

Code Metal (93/100) and Pagaya Technologies (93/100) exemplify the power of specialized, data-driven moats in highly regulated sectors. Code Metal stands out with a 95/100 in both team and growth by solving the hardest problem in defense AI: verifiable correctness. By providing mathematical proofs for AI-generated code translations, they unlock the mission-critical edge computing market that statistical LLMs cannot touch. Meanwhile, Pagaya's 94/100 in funding and growth stems from its capital-light, two-sided network moat. With $3.5T in processed loan applications, their AI credit decisioning creates a data flywheel that traditional banks simply cannot match without taking on catastrophic balance sheet risk.

We are also seeing massive signal strength in companies pioneering novel architectural approaches to existing problems. Koi (93/100) recognized that the modern enterprise threat vector isn't compiled binaries, but unvetted browser extensions and AI agents. Their agentless "Supply Chain Gateway" earned them a 95/100 in growth and culminated in a rapid $400M acquisition by Palo Alto Networks. Vega Security (90/100) tackles the "data gravity" problem in SIEM by moving the analytics to the data via federated search, eliminating six-figure ingestion fees. This architectural inversion gives them an 88/100 in differentiation and a massive structural cost advantage over legacy incumbents like Splunk.

Finally, hardware-software bridges are demonstrating exceptional signal strength. ZaiNar (92/100) achieves a 95/100 in technology by turning existing 5G/Wi-Fi networks into sub-meter 3D positioning grids without new hardware—a software-only moat protected by 90 issued patents that is perfectly timed for the autonomous robotics boom. Portkey (90/100) secures its position by owning the "AI Gateway," processing 500B+ tokens daily to provide the essential governance, routing, and observability layer that creates permanent vendor lock-in for enterprise GenAI deployments.

Sector Themes

Across the dataset, capital is aggressively concentrating on the intersection of AI and the physical world—what we term "Physical AI." Sectors like AI-Native Metal Manufacturing (Freeform), Energy/Climate AI (Emerald AI), and Enterprise Integration (Circuit) are seeing unprecedented momentum. The market has realized that software-only LLMs have reached a plateau of enterprise utility without physical grounding. Companies that can translate AI reasoning into physical action—whether it is Freeform using real-time sensor data to autonomously 3D print flawless rocket engine parts, Circuit transforming complex CAD files into actionable manufacturing workflows, or ZaiNar providing the 3D location tracking required for industrial robotics—are commanding premium valuations.

Geographically, while San Francisco remains the undisputed heavyweight champion for foundational models and spatial intelligence (OpenAI, World Labs, Cogent Security), Tel Aviv has solidified its absolute dominance in Cybersecurity AI. Both Koi and Vega Security leverage elite Unit 8200 military intelligence pedigrees to build highly defensible enterprise security infrastructure. Furthermore, we are seeing specialized vertical AI hubs emerge outside traditional coastal clusters, such as Salt Lake City for Fintech AI (Jump), Austin for Enterprise Integration (Circuit), and Freiburg, Germany for visual intelligence (Black Forest Labs). Domain expertise is increasingly decentralizing capital allocation away from generalist tech hubs.

In terms of momentum versus stagnation, agentic security and Fintech AI show massive momentum, while generic productivity wrappers are entirely stagnant. Kontigo's approach to Fintech AI perfectly illustrates this momentum: by bypassing app fatigue entirely and using WhatsApp as an AI-powered banking OS built on USDC stablecoins, they achieved $30M ARR and profitability within 12 months. This is the blueprint for modern AI distribution: intercept the user where they already operate, leverage AI for hyper-personalized service, and utilize modern infrastructure (blockchain/stablecoins) to bypass legacy bottlenecks.

Investment Outlook

Looking ahead to next week and the coming quarter, our conviction level is unequivocally "Strong Buy" for infrastructure orchestration and verifiable agentic AI. The $400M acquisition of Koi by Palo Alto Networks is a starting gun for an M&A supercycle in the agentic security space. Incumbents are desperate to plug the massive security blind spots created by developer autonomy and AI agent proliferation. We must aggressively evaluate our deal flow for "Supply Chain Firewall" architectures and autonomous remediation platforms like Cogent Security. The consolidation in this sector will be rapid and highly lucrative for early-stage backers.

Specifically, the committee should closely monitor three critical assets. First, Emerald AI: as the 2026 grid regulations bite and hyperscalers face 7-year interconnection delays, Emerald's software-based power orchestration is the most capital-efficient unlock in the market. Watch their upcoming UK live trial with National Grid; if successful, it validates a massive TAM expansion. Second, Code Metal: as defense budgets increasingly allocate toward autonomous edge computing, Code Metal's "Awardable" status in the DoD's Tradewinds marketplace positions them to capture the lion's share of legacy code modernization. Third, Vega Security: if their federated search engine successfully displaces legacy SIEM deployments by eliminating ingestion fees, they will define the next generation of Security Analytics Mesh.

The era of "AI as a feature" is over. We are entering the era of "AI as the operating system" and "AI as physical infrastructure." The barbell strategy is no longer theoretical; it is a strict mandate. We will deploy capital aggressively into companies building the foundational infrastructure of the spatial web, or the vertical operating systems governing the physical and financial worlds. Anything in the middle is a structural trap.

Data by AI Market Watch