Venture Market Update: July 23, 2026 - CuspAI Raises $450 Million at $2.6 Billion Valuation, Neo Exits Stealth with $100 Million, Deals from Natural, Empirical Security, Infinity, Brenus Pharma, and Plazza, an Analysis of Capital Shifts in AI's Control Layer for Venture Investors and Funds
Key Takeaway of the Day: The venture market has shifted its focus from paying for access to models to paying for control over the bottlenecks surrounding them. CuspAI's $450 million raise at a $2.6 billion valuation, the stealth exit of Neo with $100 million, and a wave of strategic investments from corporations and private equity are reshaping the logic of capital allocation. We analyze what these changes mean for venture funds and LPs.
Venture investments in mid-July 2026 are not widely distributed across the startup market. Capital is clearly skewing towards infrastructure, security, and software that exists in the control layer of artificial intelligence, rather than the presentation layer. The largest check of the cycle has gone into AI material development, while other notable rounds have clustered around cybersecurity, inference software, payment rails for AI agents, and automation of regulated workflows.
This combination is significant for investment committees. It indicates that funds still seek exposure to AI but increasingly prefer businesses that shape the computation economy, control data, or oversee critical corporate processes, rather than yet another thin overlay on a frontier model.
Deal of the Day: CuspAI Raises $450 Million at $2.6 Billion Valuation
The Series B round for UK-based CuspAI has become the defining transaction of the week, signaling where deep-pocketed investors see the next defensive moat in AI—not just in models, but in the physical systems that those models help design.
- Round Size: $450 million, Series B, valuation of $2.6 billion.
- Syndicate: Led by Kleiner Perkins and NEA, with participation from Bezos Expeditions, the UK government, AMD Ventures, Lux Capital, Glade Brook Capital Partners, and Invest-NL.
- Total Funding: Over $650 million in just two years since its launch.
- Headquarters: Cambridge, UK.
The company utilizes AI to discover new materials, focusing on semiconductors, batteries, clean energy, and advanced manufacturing. Investors are interested in how these materials are positioned upstream in several constrained markets. If AI can reduce semiconductor manufacturing reliance on rare metals, shorten R&D cycles, or improve energy materials, the returns extend beyond software multiples—they cascade into manufacturing economics, supply chain resilience, and geopolitical competitiveness.
For founders, the lesson here is stark: such a level of capital intensity in deep tech is funded only when the project is tied to strategic industrial demand rather than abstract scientific promise.
Cybersecurity as a Magnet for Venture Capital
The second major cluster of deals is cybersecurity, and it is no coincidence. AI is not only creating new categories of software but also rewriting the risk model for existing ones.
Neo: $100 Million Stealth Exit
Boston-based Neo has raised $100 million through a combined seed and Series A round led by Andreessen Horowitz, Bessemer Venture Partners, Craft Ventures, and Merlin Ventures. Founded by former SentinelOne executives Nick Warner and Shlomi Salem, along with technologist Eran Shirazi, their thesis is straightforward: traditional corporate security tools are poorly equipped for the world of AI applications and agent systems. The platform enables security teams to see, verify, and control AI software before access to data or automated actions create new operational risks. The technology is already undergoing pilots in finance, energy, and transport.
Empirical Security: $25 Million for Predicting Exploited Threats
The Chicago-based company raised a Series A led by Brightmind Partners, with participation from HPA and Costanoa Ventures, bringing total funding to $37 million. Their positioning is noteworthy: rather than broad rhetoric around "AI security," the company focuses on predicting threats through the monitoring of exploited vulnerabilities. Budgets are opening up faster for software that helps prioritize specific vulnerabilities than for platforms promising simply "more intelligence."
Second-Order AI Stack: Software That Makes Hardware Useful
Venture investors are paying particular attention to Infinity's $15 million seed round at a post-money valuation of $100 million. The company builds a software layer that makes any AI chip ready for inference.
The investment thesis is simple: new chips mean little if developers cannot quickly deploy on them. Nvidia's dominance in AI largely stems from software and ecosystem maturity, rather than just hardware performance. Infinity effectively sells time to usefulness: if new silicon manufacturers can become inference-ready in days instead of months or years, they have a chance to compete for manufacturing demand.
A deeper signal indicates that venture capital is taking the "second-order AI stack" seriously. The market has already spent heavily on model developers and chip companies. The next funds flow to translators, adapters, and orchestration layers that make this infrastructure usable.
Agent Commerce: Payment Rails for AI
Startup Natural has closed a Series A at $30 million led by Kirsten Green from Forerunner, bringing total funding to $40 million. The company addresses a problem that will grow with every viable agent scenario: how software conducts financial actions on behalf of a user or company without chaos in access rights, payment friction, and compliance issues.
The investor logic is clear:
- Agent commerce is easy to demonstrate and hard to scale industrially.
- Once software begins to buy software, pay suppliers, and handle transactional processes, the rails themselves become the product.
- The owner of this layer captures volume, compliance, and built-in distribution far beyond the capabilities of a thin application.
This gives the company a more sustainable position than many applied AI startups, whose differentiation erodes as underlying models improve. For founders, the distinction is crucial: AI that saves a click will struggle to attract; AI that safely moves a dollar attracts strategic capital.
The Return of Strategic Capital: PE, Corporations, and Distribution Channels
One of the most telling features of the current market is the noticeable share of strategically important financings that have come not from classic venture funds, but from private equity, corporate, and ecosystem partners.
- Quorum (Washington) received undisclosed strategic investment from Enlightenment Capital. The AI-based platform for government affairs serves over 2000 organizations, including more than half of the Fortune 100 companies. The capital will be directed toward executing the product roadmap and expanding agent AI capabilities.
- Wagmo (New York) secured strategic investment from Curql to take modern veterinary healthcare insurance to credit union channels—a prime example of distribution-oriented capital.
- HALO X-ray Technologies (Nottingham, UK) closed a multimillion-pound round led by Agilent, with participation from the UK Innovation Science Seed Fund and Midland Engine Investment Fund to complete regulatory approval of X-ray diffraction technology in screening systems.
When buyers, channels, or industry experts can fund part of the next chapter of growth, founders become less dependent on purely financial sponsors. In a tighter capital market, this is an advantage.
Biotech and Healthcare: Financing for Milestones Rather Than Narratives
Lyon-based Brenus Pharma added €11 million to its Series A, bringing total funding since inception to €38 million. This expansion is tied to achieving clinical, regulatory, and business development milestones surrounding STC-1010—the leading clinical immunotherapy program against stomach and colorectal cancer. The company specifically noted the arrival of new investors from Europe and the Asia-Pacific region.
Such expansions are important as an indicator of risk underwriting. Instead of forcing every company into a new narrative reset, investors are willing to add capital when the team has sufficiently de-risked the science. This is often healthier than a completely new round at an inflated valuation, as it directly ties capital to progress.
In India, Plazza (Bengaluru) raised $15 million Series A led by Accel, Elevation Capital, and Nexus Venture Partners for the expansion of its pharmacy network and instant delivery of medications. This is a bet on logistics and trust in a category where reliability outweighs branded storytelling: availability, order fulfillment rates, inventory routing, and area coverage density form a true defensive moat.
Capital Geography: A Market Without a Single Template
The current venture pipeline is geographically mixed but uneven:
- The USA dominates early-stage software and cybersecurity—Neo, Empirical Security, Infinity.
- The UK has captured the largest check of the cycle through CuspAI and demonstrated strength in deep tech with government capital participation.
- France has emerged in biotech and clinically validated assets.
- India has entered the agenda through operationally dense commerce in healthcare, rather than frontier AI.
Global venture does not converge on a single template. Different regions attract capital where they already have talent density, regulatory competency, or operational advantages.
Cycle Risks: Where Venture Funds May Overpay
The discipline of the current market does not exempt portfolios from structural threats:
- Risk of Commoditization. AI applications built on widely available models may grow, but sustainable money is shifting under or around the model layer.
- Inequity in Information Disclosure. A significant portion of strategically interesting transactions takes place without disclosed sums, making benchmarking valuations difficult.
- Capital Intensity of Deep Tech. Computations, laboratory processes, and industrial partnerships are costly, meaning constant dilution of early investors' shares.
- Concentration in Narrow Categories. When the market pays primarily for infrastructure, security, and science, risk correlation within the portfolio increases.
- Dependency on Regulatory Milestones. In biotech and physical security, approval timelines remain a primary source of uncertainty.
Conclusions for Venture Investors and Funds
The current deal flow demonstrates a market attempting to assess not novelty but where AI creates sustainable scarcity. In some cases, this is scarce scientific competence, as seen with CuspAI. In others, it is scarce trust, as seen in cybersecurity and public policy software. In third instances, it is scarce operational reliability, as characterized by medication delivery.
Practical takeaways for investment committees:
- Invest in Bottlenecks, Not Slogans. If a startup touches on infrastructure cost, security posture, compliance processes, or long-term order fulfillment, large rounds are still underwritten.
- Seek Accumulating Defensibility. Scientific intellectual property and industrial partnerships, founder reputation, ecosystem leverage, and progress on scientific milestones—the common denominator is not technology, but the ability to make a replacement painful.
- Consider Investor Type as a Value Factor. Sometimes the most valuable investor is not the one paying the highest price, but the one opening the cheapest and most secure path to customers.
- Ask What Changes with the Next Dollar. Expansions, strategic investments, and concentrated early rounds are displacing broad syndication based on hype—both sides of the market are becoming more disciplined.
- Bet on the Layers around Autonomy. Payments, security, chips, science, and workflow infrastructure benefit from AI while remaining difficult to commoditize. This is where premium multiples are likely to concentrate.
The next phase of startup financing looks less like a race to slap AI onto everything and more like a competition for ownership of systems that make AI safe, deployable, and economically viable. Companies winning capital now are not just promising automation—they are defining who controls the bottlenecks around it.
