← Back to Blog

Agentic AI Deals: M&A, Acquisitions, and VC Funding (July 24-31, 2026)

Anthropic advances toward an October IPO with secondary markets implying a valuation above 1 trillion dollars on annualized revenue past 44 billion dollars, Moonshot AI closes 3.5 billion dollars at a 35 billion valuation on Kimi K3, NVIDIA weighs a 250 billion dollar guarantee inside a 750 billion dollar OpenAI data center package, Cognition acquires Poke while Midjourney and World Labs also buy, Helsing raises 1.8 billion dollars for defense AI, Nous Research raises at a 1.5 billion valuation, and cybersecurity consolidation continues as Palo Alto Networks closes CyberArk and hyperscalers commit roughly 700 billion dollars in 2026 AI capex.

📅 July 30, 2026 ✍️ Authored by AI ⏱️ 15 min read 📁 AI

The final week of July 2026 underscored that capital is consolidating around the agent infrastructure stack at a pace and scale without precedent, spanning frontier labs, gateway and identity security, defense, and the data center buildout that underpins all of it. Anthropic advanced toward an October IPO with secondary markets pricing the company above 1 trillion dollars, a premium over the 965 billion dollar post-money valuation set in its 65 billion dollar May Series H, on annualized revenue that leapt from 9 billion dollars at the end of 2025 to more than 44 billion dollars by May 2026. In China, Moonshot AI closed a larger-than-targeted 3.5 billion dollar round at a 35 billion dollar valuation on the strength of its Kimi K3 model. NVIDIA was reported to be weighing a 250 billion dollar financial guarantee inside an OpenAI data center package whose contours exceed 750 billion dollars. On the M&A side, Cognition acquired the consumer assistant Poke while Midjourney and World Labs also announced acquisitions, and cybersecurity consolidation continued its 2026 run as Palo Alto Networks closed CyberArk, CrowdStrike integrated Seraphic Security and its 740 million dollar SGNL deal, and Zscaler absorbed SquareX. Defense AI leader Helsing raised 1.8 billion dollars led by JPMorgan Chase, Lightspeed, and Iconiq, and Nous Research raised at a 1.5 billion dollar valuation. With the four largest hyperscalers on track to spend roughly 700 billion dollars in 2026 AI capex, the week made clear that the money is flowing toward whoever controls the model, the gateway, and the compute beneath both.

Anthropic Nears a Trillion-Dollar Public Debut

Anthropic moved decisively toward an October IPO, having filed a confidential S-1 with the SEC, with secondary-market trading implying a valuation between 1.05 trillion and 1.15 trillion dollars, a clear premium over the 965 billion dollar post-money mark set when it closed a 65 billion dollar Series H in May 2026, the largest private AI funding round in history. The valuation is underpinned by extraordinary revenue growth, with annualized revenue rising from roughly 9 billion dollars at the end of 2025 to more than 44 billion dollars by May 2026, a more than fivefold increase in under six months, driven heavily by coding and agentic workloads. The launch of Claude Opus 5 this week at half the price of Claude Fable 5, together with a multi-year AMD compute partnership, reinforced the growth and supply-resilience story Anthropic is taking to public investors.

An Anthropic listing at or above a trillion dollars would be the first AI-native company to debut at that scale and would set a public-market benchmark against which every other lab is valued. It would also give Anthropic a currency for acquisitions and a capital base to fund the compute commitments that frontier development demands.

Why it matters: Anthropic's trajectory reframes the entire AI capital market. By overtaking OpenAI's 852 billion dollar valuation on the private unicorn board and approaching a public debut above a trillion dollars, Anthropic has demonstrated that agentic and coding revenue can compound faster than almost any prior software category. That validates the thesis that value accrues to whoever owns the model layer that agents run on, and it pressures every competitor to match both capability and price.

The ripple effects reach directly into the security and gateway markets covered across these blogs. As enterprises standardize on Claude through AI gateways, the vendors that control those gateways, from Palo Alto Networks Prisma AIRS to Cloudflare, Kong, and ServiceNow AI Gateway, gain leverage over the traffic Anthropic monetizes. A trillion-dollar Anthropic also intensifies the compute arms race, feeding the same hyperscaler capex and data center financing that dominate the rest of this week's deal flow, and it raises the stakes for the identity and governance layers that must sit between autonomous agents and the models they consume.

Moonshot AI Closes 3.5 Billion at a 35 Billion Valuation

Moonshot AI, the Beijing-based frontier lab, closed a 3.5 billion dollar financing round at a 35 billion dollar valuation, far exceeding the 1 billion to 2 billion dollars it had initially targeted, on the momentum of its breakthrough Kimi K3 model that drew attention across Silicon Valley. The round marks one of the largest Chinese AI raises of the year and signals that the frontier model race is not confined to US labs, with Moonshot positioning Kimi K3 as competitive with Western flagship models at a fraction of the training cost.

The scale of the oversubscription, more than double the top of the initial target, reflects intense investor appetite for a credible non-US frontier contender, particularly one whose model performance has forced comparison with the latest releases from OpenAI, Anthropic, and Google.

Why it matters: Moonshot's raise establishes a well-capitalized frontier competitor outside the US regulatory and export-control perimeter, which has direct implications for the multi-model reality that enterprises and gateways must manage. As Kimi K3 and its successors reach global developers, the number of frontier providers that an AI gateway must route, govern, and secure grows, reinforcing the strategic value of the control plane over any single model.

The geopolitical dimension also matters for the security vendors tracked here. A Chinese frontier lab at a 35 billion dollar valuation sharpens the sovereignty and data-governance questions that drive demand for gateway-level controls over which models an enterprise is permitted to call. IBM's Sovereign Core, ServiceNow AI Control Tower, and Check Point's AI Defense Plane all address the need to enforce policy over model selection, and a stronger field of non-US models makes that enforcement a board-level concern rather than a technical preference.

NVIDIA Weighs a 250 Billion Guarantee for OpenAI Compute

Reports surfaced this week that NVIDIA was considering a financial guarantee of as much as 250 billion dollars to backstop OpenAI's massive data center buildout, with the broader contours of the arrangement totaling more than 750 billion dollars across compute, financing, and infrastructure commitments. The structure would tie NVIDIA's balance sheet directly to OpenAI's capacity expansion, an unprecedented vendor-financing arrangement that reflects both the scale of compute OpenAI requires and NVIDIA's incentive to secure its largest customer's growth.

The reported package sits alongside OpenAI's 852 billion dollar valuation following its 122 billion dollar March round and its move this week to open GPT-5.6 to general availability after clearing a US government review. Financing at this magnitude blurs the line between supplier and investor and concentrates the AI supply chain around a small number of interdependent players.

Why it matters: A 250 billion dollar guarantee inside a 750 billion dollar package is a signal that compute has become the binding constraint on frontier AI, and that the capital required to relieve it now exceeds what even the largest labs can raise through conventional equity. This is the structural backdrop to Anthropic's AMD partnership and its trillion-dollar IPO, both of which are ultimately about securing and financing compute.

The concentration risk is significant for the broader ecosystem. When a single vendor guarantees a single customer's expansion at this scale, the entire agentic economy inherits exposure to that relationship. For security and gateway vendors, the practical consequence is that inference and agent traffic will continue to explode as capacity comes online, which is precisely the volume that Fortinet's Intel Security Processor 6, Palo Alto Networks Prisma AIRS, and every other gateway must be built to inspect. The compute buildout and the security control plane are two sides of the same expansion.

Cybersecurity Consolidation Continues Through 2026

The cybersecurity M&A wave that defined the first half of 2026 carried into the summer, driven by platform consolidation, AI disruption, and hyperscaler appetite. Palo Alto Networks completed its acquisition of identity-security vendor CyberArk in February 2026, adding a cornerstone to its identity-first agentic security strategy, and its total 2026 AI-related acquisition spend now exceeds 700 million dollars across the Portkey gateway and identity assets. CrowdStrike opened the year with its January acquisition of Seraphic Security for browser security and its 740 million dollar purchase of identity startup SGNL, while Zscaler acquired SquareX in February, kicking off a browser-security arms race among the platform leaders.

OpenAI's Trusted Access for Cyber program, launched in April 2026, effectively anointed Palo Alto Networks, CrowdStrike, and Zscaler as platform consolidators, squeezing sub-scale pure-play vendors. Analysts note that the top ten cybersecurity vendors each hold only one to two percent market share, but the gap between platform leaders and point-solution providers is widening rapidly as enterprises pay a premium to reduce vendor count.

Why it matters: The consolidation is a direct expression of the platformization thesis that Palo Alto Networks CEO Nikesh Arora coined, and it is reshaping who can compete in agentic security. By acquiring CyberArk for identity and Portkey for the gateway, Palo Alto has assembled the most complete control-plane stack of any vendor, spanning identity, gateway, and runtime, which is why it commands a premium in the M&A market.

The competitive responses diverge sharply. CrowdStrike is building through identity and endpoint acquisitions, Zscaler through browser and edge security, and Check Point through its 300 million dollar Lakera prompt-injection buy and its Google agent-fabric partnership. Fortinet, notably, is resisting the acquisition path in favor of organic development and its Intel Security Processor 6 hardware collaboration. The strategic question for enterprises is whether the agentic security future belongs to the acquirer that assembles the widest platform or the builder that controls its own silicon and code, and this week's deal flow suggests the market is still funding both bets simultaneously.

Cognition, Midjourney, and World Labs Go Shopping

The week saw a cluster of acquisitions by AI-native winners deploying their capital and equity to expand. Cognition, the AI coding company, acquired Poke, an AI assistant that texts users conversationally with humor and slang and had processed 100 million messages in its prior three months, extending Cognition's reach from developer tooling toward consumer engagement. Midjourney and World Labs also announced acquisitions during the same week, part of a broader pattern that Forbes characterized as AI's early winners shopping for their next big moves. The acquisitions reflect a maturing market in which the best-funded AI startups are becoming acquirers rather than targets.

These deals arrive against a record-setting backdrop. Q2 2026 was the biggest quarter on record for billion-dollar acquisitions, with 24 companies bought at or above 1 billion dollars for 113 billion dollars in total value, alongside 32 venture-backed IPOs above 1 billion dollars, and the headline transaction was SpaceX buying Anysphere, the maker of Cursor, for 60 billion dollars in all stock on June 16, the largest acquisition of a venture-backed startup ever.

Why it matters: The shift of AI-native companies into the acquirer role marks a new phase of the cycle in which value created by frontier tooling is being recycled into consolidation. Cognition's move into consumer assistants and the Anysphere Cursor deal both show coding and agent companies extending horizontally, which multiplies the number of agent surfaces that enterprises will eventually need to govern.

For the security and gateway landscape, this consolidation among AI-native firms mirrors the consolidation among security platforms. Just as Palo Alto Networks and CrowdStrike are assembling control-plane stacks, coding and agent companies are assembling capability stacks, and every acquisition adds new integrations, new tools, and new MCP surfaces. The Cursor acquisition by SpaceX, feeding directly into xAI Grok 4.5 this week, is the clearest example of how model, tooling, and compute are converging inside single corporate entities, which raises the strategic premium on the neutral gateway layer that can govern across all of them.

Defense AI and Emerging Labs Draw Fresh Capital

Capital continued to flow toward defense-oriented and independent AI labs. Helsing, the European defense AI company, raised 1.8 billion dollars in a round led by JPMorgan Chase, Lightspeed Venture Partners, and Iconiq, announced July 13, cementing its position as the continent's most valuable defense AI firm amid heightened government demand for autonomous and AI-enabled capabilities. Nous Research raised at least 75 million dollars at a 1.5 billion dollar valuation in a round led by Robot Ventures, also announced July 13, reflecting continued investor appetite for open and decentralized model development. These raises sit within a record first half, as North American startup funding shattered records in the first half of 2026, driven overwhelmingly by AI.

The breadth of funding, from trillion-dollar frontier labs to billion-dollar defense specialists to smaller open-research outfits, shows capital spreading across the full spectrum of the AI ecosystem rather than concentrating solely at the top.

Why it matters: The Helsing raise signals that AI investment is expanding beyond commercial software into defense and sovereign applications, a category with distinct security and governance requirements that will drive demand for the most rigorous forms of agent control. Defense deployments cannot tolerate the prompt-injection and gateway-compromise risks that dominate this week's security roundup, which positions vendors like Palo Alto Networks, Fortinet, and IBM with government credentials to capture that spending.

The persistence of funding for independent labs like Nous Research also matters for the multi-model future. Every new model provider adds an endpoint that enterprises must inventory and govern, reinforcing the value of the gateway and identity control planes. As capital continues to fund a widening field of model providers, from Moonshot AI in China to Nous Research in open development, the arbitration layer that routes and secures traffic across all of them becomes the most defensible position in the stack, which is why security vendors are willing to pay platform premiums to own it.

Hyperscaler Capex Anchors the Deal Economy

Beneath the headline financings sits the largest capital commitment of the cycle: the four dominant hyperscalers, Alphabet, Amazon, Microsoft, and Meta, are on track to spend roughly 700 billion dollars in 2026 AI capital expenditure, funding the data centers, accelerators, and networking that every model and agent ultimately runs on. That figure dwarfs the individual funding rounds and acquisitions of the week and represents the demand signal behind NVIDIA's reported 250 billion dollar guarantee for OpenAI and behind the vendor-financing arrangements now knitting suppliers and customers together. The capex is not discretionary; it is the price of remaining competitive as inference and agent workloads scale, and it is increasingly financed through debt, guarantees, and off-balance-sheet structures rather than cash alone.

The scale of the spend is reshaping adjacent markets, from power and cooling to custom silicon, and it is the reason Fortinet partnered with Intel on Security Processor 6 to handle AI data center throughput and why every gateway vendor is architecting for volumes that did not exist a year ago.

Why it matters: A 700 billion dollar annual capex base establishes the AI buildout as one of the largest infrastructure investments in corporate history, and it locks the hyperscalers into multi-year commitments that they must monetize through cloud, model, and agent-platform revenue. That monetization runs directly through the agent fabrics each hyperscaler is building, from Google Agent Gateway to AWS Bedrock AgentCore to Microsoft Agent 365, which are the commercial return on the capex.

For security vendors, the capex wave is both an opportunity and a competitive threat. The opportunity is the explosion of agent traffic that must be inspected and governed. The threat is that the hyperscalers, having spent 700 billion dollars, will bundle their own governance into their agent platforms, as Check Point is betting by partnering with Google rather than competing with it. Whether independent gateway vendors like Palo Alto Networks, Cloudflare, and Kong can hold their ground against hyperscaler-native governance is the defining commercial question that this capex creates, and it is why control of the gateway remains the most contested position in the stack.

The Exit Window Reopens

The 2026 deal economy is being reinforced by a reopened exit window that gives investors and employees liquidity and validates late-stage valuations. Q2 2026 saw 32 venture-backed IPOs above 1 billion dollars alongside its 24 billion-dollar acquisitions, and Anthropic's confidential S-1 and targeted October listing would extend that momentum into the largest AI IPO yet. The combination of a robust IPO market and record M&A gives founders genuine optionality between going public and selling, which in turn sustains the private funding rounds that dominate this week's flow by giving investors credible paths to return.

A healthy exit environment also accelerates consolidation, because acquirers with public currency and IPO-bound companies with fresh capital both have the means to buy. The cybersecurity consolidation led by Palo Alto Networks, CrowdStrike, and Zscaler is a direct beneficiary of this dynamic.

Why it matters: A functioning exit window is what separates a durable investment cycle from a bubble, because it lets capital recycle rather than remain trapped in illiquid positions. The 32 billion-dollar IPOs and 24 billion-dollar acquisitions of Q2 signal that the market can absorb AI exits at scale, which reduces the risk that the record funding of the first half becomes stranded.

For the agentic ecosystem specifically, a strong exit environment funds the next generation of infrastructure and security startups even as the current leaders consolidate. The AI gateway startups tracked across these blogs, from Portkey, now inside Palo Alto, to Kong and LiteLLM, exist because early capital saw a path to exit, and Anthropic's potential trillion-dollar debut will only strengthen that conviction. The capital consolidating around the agent infrastructure stack this week is, in large part, capital that expects to exit through the same window Anthropic is about to open.

Numbers at a glance

Anthropic approached an October IPO priced between 1.05 trillion and 1.15 trillion dollars on secondary markets, above its 965 billion dollar May Series H that raised 65 billion dollars, on annualized revenue past 44 billion dollars. Moonshot AI raised 3.5 billion dollars at a 35 billion dollar valuation. NVIDIA weighed a 250 billion dollar guarantee inside a 750 billion dollar OpenAI package, against OpenAI's 852 billion dollar valuation. SpaceX bought Anysphere for 60 billion dollars, the largest venture-backed acquisition ever, part of a record Q2 with 24 billion-dollar acquisitions worth 113 billion dollars and 32 billion-dollar IPOs. Helsing raised 1.8 billion dollars, Nous Research raised at a 1.5 billion dollar valuation, CrowdStrike paid 740 million dollars for SGNL, Palo Alto Networks exceeded 700 million dollars in AI acquisitions, and hyperscalers are on track for roughly 700 billion dollars in 2026 capex.