Headline & intro
Google Cloud Next 2026 was marketed as a celebration of AI innovation, but the headline move was bluntly financial: $750 million to push AI agents into enterprises via partners. On stage it looks like generosity toward startups. In practice, it’s Google’s most aggressive attempt yet to hard‑wire the next generation of AI‑native companies to its infrastructure.
This piece looks at who actually wins from that $750 million, how the highlighted startups fit Google’s strategy, what it means for competitors, and why European founders and CIOs should read the small print before cashing the cheques.
The news in brief
According to TechCrunch’s coverage of Google Cloud Next 2026 in Las Vegas, Google announced a new $750 million budget dedicated to partners selling AI agents to enterprise customers. The money is available to a wide range of partners, from early‑stage startups to large consulting firms, and can be used for proof‑of‑concepts with Gemini models, access to Google engineers, cloud credits and deployment incentives.
TechCrunch notes that Google also showcased a long roster of startups building on its cloud. Standout names included Lovable, which is rolling out a new coding agent via Google’s enterprise app marketplace and was tracking toward roughly $400 million in annual recurring revenue earlier this year, and Notion, the $11 billion‑valued productivity platform using Gemini models for text and image features.
Other highlighted companies ranged from Gamma (an AI‑driven presentation tool using Google’s Nano Banana 2 image model) and open‑source projects like ComfyUI and Inferact, to vertical AI players in logistics, healthcare, insurance, hospitality and sustainability.
Why this matters
On paper, $750 million for partners looks like Google writing checks to help startups. Strategically, it is Google paying to buy time and gravity in the most contested part of the AI stack: where agents meet enterprise workflows.
The winners in the short term are clear:
- Google Cloud gains a powerful sales amplifier. Every euro of credits or co‑funded PoCs is designed to turn into long‑term consumption of compute, storage and Gemini APIs.
- AI‑native startups like Lovable, Gamma, Inferact or Vapi get subsidised infrastructure and go‑to‑market support at the exact moment when GPU costs and customer acquisition are painful bottlenecks.
- Consultancies and systems integrators can pitch ambitious AI agent projects to corporate clients with Google footing part of the bill.
But there are losers and risks:
- Competing clouds (AWS, Azure and smaller players) now face a Google funded land‑grab for the most valuable early adopters.
- Startups risk building on heavily discounted infrastructure that later becomes expensive to escape. The economics can flip the moment credits or rebates expire.
- Enterprises may end up with critical workflows mediated by partner‑built agents that are deeply tied to one cloud, making future negotiations asymmetric.
This program isn’t charity; it is customer acquisition at hyperscaler scale. The startups TechCrunch lists form a carefully curated portfolio: coding copilots, document tools, developer platforms, logistics, healthcare, insurance and sustainability – exactly the categories where AI agents are closest to becoming recurring enterprise spend.
The bigger picture
This move sits at the intersection of three broader industry trends.
1. The AI stack is shifting from models to agents.
Foundation models are commoditising faster than expected. Open‑source projects (like vLLM, from which Inferact emerged) and a flood of model providers are driving prices down. The battleground is moving to orchestration and agents that perform real tasks inside CRM, ERP, security tools and custom line‑of‑business apps. Google’s funding is targeted precisely at teams building those last‑mile agents.
2. Hyperscalers are racing to become the default home for AI‑native startups.
Microsoft has OpenAI and a well‑oiled Azure credit machine. AWS has Bedrock, Trainium/Inferentia and similar partner funds. Google came late to the cloud market; it cannot afford to be late to AI‑native workloads. Backing companies like Lovable, Notion, Gamma or ComfyUI is a way to signal that “the cool AI companies run on Google Cloud” – the same playbook AWS once used with early unicorns.
3. Open source and proprietary are converging in the cloud.
Projects such as ComfyUI and the vLLM ecosystem highlight a pragmatic reality: even the most open tools often depend on proprietary GPU capacity from hyperscalers. Google is happy to be the place where open‑source inference becomes commercially viable, as long as it captures the margins on GPUs and managed services.
Historically, this resembles the early smartphone era: app developers got visibility, tooling and revenue; Apple and Google got 30% and platform dominance. Today’s AI startups are the app developers of the cloud era – except this time, the tax is measured in compute spend and data gravity, not app‑store fees.
The European / regional angle
For European founders and CIOs, this announcement is both an opportunity and a warning.
On the upside, EU startups building AI agents in regulated sectors – think Proximal Health in insurance, ExaCare AI in post‑acute care, Watershed in sustainability reporting or logistics players like ChorusView and Stord – can potentially tap Google’s budget to co‑fund pilots with cautious European enterprises. That lowers the barrier for experimentation in markets that are traditionally conservative about new software.
However, Europe operates under GDPR, the Digital Services Act and, crucially, the emerging EU AI Act. Many of the use cases highlighted (healthcare, insurance, HR‑adjacent workflows, risk scoring) are likely to fall into “high‑risk” categories under the AI Act, carrying heavy documentation and oversight requirements.
Building these systems on a US hyperscaler raises familiar sovereignty questions:
- Where is data stored and processed?
- How easy is it to audit and explain the behaviour of Gemini‑powered agents?
- Can a company move models and data if it later prefers a European cloud or needs a sovereign setup for public‑sector work?
European cloud providers and initiatives (from OVHcloud and Scaleway to Deutsche Telekom’s offerings and various sovereign‑cloud projects) will portray Google’s $750 million as proof that AI agents are strategically decisive – and argue that Europe needs its own stack, not just subsidised access to US infrastructure.
For EU enterprises, the smart move is to treat these Google‑funded pilots as experiments, not as irreversible commitments, and to demand strict portability and data‑residency guarantees from day one.
Looking ahead
The immediate effect of Google’s program will be a wave of “built on Google Cloud” labels in AI startup pitch decks and more aggressive co‑selling into large enterprises. Over the next 12–24 months, expect three developments.
1. A subsidy war for AI startups.
AWS and Microsoft will not watch passively while Google signs up high‑potential AI‑native companies. Expect matching or larger partner funds, even more generous credits, and special terms for promising open‑source ecosystems. Startups will be tempted to multi‑cloud purely for subsidy arbitrage.
2. Rising importance of architecture decisions.
Founders will need to think beyond the next fundraising round. Are their agents tightly coupled to Gemini, or can they swap in other models? Is business logic portable, perhaps via open‑source frameworks like vLLM or standardized agent protocols, or is it deeply baked into one provider’s PaaS services? The cost of a wrong answer may only surface in three to five years, when renegotiating enterprise contracts.
3. Enterprises push back on opaque AI costs.
CIOs burned by previous cloud bills will demand clearer TCO models for AI agents. They will ask partners showcasing Google‑backed solutions to demonstrate not only accuracy and productivity gains but also predictable long‑term economics.
Unanswered questions remain. How much of the $750 million is true incremental funding versus re‑labelled marketing budget? Will Google tilt the playing field towards partners who commit to exclusivity? And how will regulators view large hyperscalers using financial muscle to shape AI markets that they also supply with critical infrastructure?
For European policymakers watching the EU AI Act move from text to enforcement, this kind of program may become a test case for how industrial policy and digital sovereignty collide with global cloud economics.
The bottom line
Google’s $750 million push for AI agents via startups is smart, aggressive and anything but neutral. It will accelerate useful experimentation, especially in verticals like healthcare, logistics and insurance, and many founders would be irrational not to take the money. But it also deepens dependency on a single cloud layer that is becoming as strategic as the operating system once was.
Before celebrating the free credits, founders and European CIOs should ask a simple question: what will it cost us to leave, once the party is over?
