📍 Gradium extended its seed funding to $100 million and welcomed NVIDIA as a new investor, while Ollama raised $65 million for open-model developer distribution and Mercor moved to acquire Deeptune’s agent-training environments. The signal is not one more AI funding cycle. Capital is moving toward the layer where models become usable work, and that changes what founders need to build, price, and defend. ───── Hey Builders! The AI market is moving from model access to operating access. The last cycle rewarded teams that could wrap a frontier model in a useful interface. The next one is rewarding companies that control where AI touches real work. That is why a Paris voice AI startup can raise a $100 million seed round seven months after launch. It is why a local open-model tool can become a venture-backed platform with 8.9 million monthly developers. It is why AI training companies are buying simulated workplaces rather than only hiring more labelers. The commercial question is becoming more specific. Who owns the voice layer? Who owns the developer runtime? Who owns the training environment? Who owns the trust signal when an agent makes decisions inside enterprise software? Ollama CEO Jeffrey Morgan put the shift plainly in the company’s funding announcement. “The future of AI is open models running everywhere work gets done.” That line matters because it captures the real market move. AI is no longer only a destination product. It is becoming a distributed operating layer. For founders, that raises the bar. It is not enough to say a product uses AI. The stronger question is whether the company controls a scarce workflow, proprietary feedback loop, trusted interface, or distribution point that makes AI more useful every day. ───── 1️⃣ $100 Million Puts Voice AI Back at the Center of the Agent Stack Gradium announced on July 8 that it extended its seed funding to $100 million, welcoming new investors including NVIDIA and opening a San Francisco Bay Area office. The company says the capital will support research, product development, international expansion, and scaling its real-time voice AI. The round is significant because voice is becoming a practical interface problem, not just a media feature. Gradium says it builds streaming speech-to-text, expressive text-to-speech, translation, and conversational intelligence for developers and enterprises that need low-latency voice experiences. TechCrunch reported that Gradium launched out of stealth in December with $70 million from FirstMark Capital, Eurazeo, DST Global Partners, Eric Schmidt, and Xavier Niel, and that the company has named Renault among its customers. The company was spun out of Kyutai and was founded by Neil Zeghidour, Laurent Mazaré, Olivier Teboul, and Alexandre Défossez. 💡 For Founders Voice will be judged by latency, reliability, privacy, and integration depth. If your product depends on natural interaction, treat voice as infrastructure and workflow design, not as a shiny front end. ───── 2️⃣ $65 Million Turns Open Models Into a Distribution Business Ollama announced a $65 million Series B led by Theory Ventures, with participation from Benchmark, 8VC, Y Combinator, Pace Capital, 49 Palms, GTMFund, and other investors. The round brings total funding to $88 million. The more important number is usage. Ollama says 8.9 million developers use the platform monthly, usage has doubled since January, the ecosystem includes more than 67,000 integrations, and the tool is used within 85% of the Fortune 500. That traction makes open-model distribution a serious software position. Ollama’s pitch is that developers can run models locally when hardware is sufficient, then scale to larger open models in the cloud without changing the basic experience. 💡 For Founders The open-versus-closed model decision is becoming an operating-design question. Build a routing strategy now so high-volume, lower-risk work can move to cheaper or more private model paths while critical reasoning stays on the strongest systems. ───── 3️⃣ 5 Million Experts Show Where Agent Training Is Really Bottlenecked Mercor announced that it will acquire Deeptune, a company building environments for reinforcement learning. Financial terms were not disclosed. Mercor’s own explanation is direct. Its network of more than five million domain experts creates tasks and verifiers, while Deeptune builds software environments where agents can practice work inside realistic replicas of enterprise applications. The company says Deeptune has recreated hundreds of enterprise applications over the past two years, from spreadsheets to Salesforce, and that Mercor was already a Deeptune customer. Fortune also reported that Mercor CEO Brendan Foody personally invested in Deeptune’s $43 million Series A before the acquisition. 💡 For Founders If agents are part of your roadmap, define the training environment before you define the launch claim. The product only becomes credible when the agent can practice, fail, and be measured against work that looks like the customer’s real workflow. ───── 4️⃣ $100 Million Makes Fundraising Itself an Agent Workflow TechCrunch reported that Lyzr, an enterprise AI agent startup, used its own agent to help run a $100 million Series B process at a roughly $500 million valuation. The system reportedly handled investor questions from more than 130 investors, drafted investment memos, and tracked which slides investors spent time on. TNW reported that the effort drew about $400 million in interest from Silicon Valley funds, Middle Eastern venture firms, and financial-sector backers. The same report noted that Lyzr’s earlier Series A was led by Rocketship.vc, with Accenture Ventures among the backers. This is not proof that founders can automate trust. It is proof that diligence, materials, routing, and repetitive investor interaction are becoming structured workflows that software can absorb. 💡 For Founders Do not confuse an agent-led process with a relationship-free process. Use agents to make your data room, metrics, FAQs, and follow-up cleaner, then keep humans focused on judgment, trust, and terms. ───── 5️⃣ 101,743 AI-Cited Cuts Put Workforce Design Back in the Brief Challenger, Gray & Christmas reported that U.S.-based employers announced 45,849 job cuts in June, down 53% from May. Artificial intelligence led all reasons for cuts for the fourth consecutive month, with 14,029 announced cuts in June. So far in 2026, Challenger says AI has been cited in 101,743 job-cut announcements, about 23% of all cuts. Technology led all sectors in June with 15,503 cuts and has announced 139,156 cuts this year, up 83% from the same period in 2025. The same report said employers announced plans to hire 10,933 workers in June, down 44% from May but above the 3,191 plans announced in June 2025. The labor market is not simply collapsing under AI. It is reallocating around new capabilities, new budgets, and new operating models. 💡 For Founders Make workforce change explicit in the product case. Buyers need to know whether your AI reduces headcount, redirects labor, improves throughput, or creates new roles, because each answer creates a different budget path and risk discussion. ───── 🔧 Three moves to make this week 1️⃣ Map your operating layer Identify the exact place where your product touches real work. Voice, runtime, training environment, data loop, or workflow verification can all be defensible if you own the customer’s repeated behavior. 2️⃣ Build the routing model Separate work that needs frontier reasoning from work that needs low cost, privacy, or local execution. This will help you price more intelligently and avoid building a product with one expensive model path for every task. 3️⃣ Turn proof into infrastructure Replace demo language with measured operating evidence. Track latency, completion rate, task quality, error recovery, and human review cost so customers can see how the system performs when work gets messy. ───── 💬 Quote of the Day "The future of AI is open models running everywhere work gets done." — Jeffrey Morgan, CEO and Co-Founder of Ollama ───── 🏁 Stay Ahead With OHUBNext This brief is part of what OHUBNext members get every day. For $5.99/month — or $59/year — you get the full daily brief, access to OHUB's Library of Opportunity, self-paced certificates in High-Growth Company Building and Tech Ecosystem Investing, career accelerator tools, and live monthly labs with the OHUB team. The annual plan includes exclusive masterclasses, early course access, and wealth tools built for builders who are serious about owning what comes next. Institutional-grade. Fraction of the cost. 🚀 Join at opportunityhub.co/next ───── 🎬 Closing Thought The strongest AI companies in this moment are not only chasing better outputs. They are building the places where models meet the world. That can look like a voice interface that responds without awkward delay. It can look like a developer runtime that makes open models practical inside companies. It can look like a simulated enterprise application where an agent learns how to work before a customer ever trusts it with production access. This is where AI becomes less abstract and more commercial. The market is asking which companies can turn capability into repeatable work, and which founders understand the difference between a demo and an operating system. For builders, the assignment is clear. Own the workflow where the model becomes useful, then measure it until the buyer can trust it. ───── ⚡️ OHUBNext Daily Brief - investments, edge tech, and moves that matter. For 12+ years, OHUB has been building pathways and on-ramps to multi-generational wealth without reliance on pre-existing wealth. Through exposure, skills, entrepreneurship, capital markets, and inclusive ecosystems, we've helped people create new jobs, new companies, and new wealth.