Founder Insights: Weekend Edition – Issue #6

The theme this week is money. The hype cycle is maturing into hard economics. Every major player is putting real numbers behind their bets, and the scale of those commitments shows where conviction lives. OpenAI’s seven-year, $38 billion cloud deal with AWS and Microsoft’s $9.7 billion pact with IREN signal that AI infrastructure is no longer just a procurement line; it is a capital market. AWS, Microsoft, and Lambda are locking in compute the way manufacturers secure steel and power, while firms like Cipher Mining are converting long-term leases into recurring AI revenue streams. The supply race is becoming a balance sheet race.

Inside the enterprise, the conversation is shifting from adoption to return. Anthropic raised its internal targets to $70 billion in 2028 revenue and $17 billion in cash flow, while OpenAI crossed the 1 million business customer mark. Palantir’s U.S. commercial revenue more than doubled as companies started paying for measurable gains in efficiency, not pilots. SAP launched its first business-tuned AI model, and Cognizant is deploying Claude to 350,000 employees, signaling that AI at work now lives inside operations at scale.

Capital flows are reinforcing the same reality. Armis raised $435 million pre IPO at a $6.1 billion valuation, MoEngage pulled in $100 million from Goldman Sachs and A91 Partners, and UnifyApps secured $50 million to build an operating system for AI workflows. Investors are backing the picks, shovels, and proven revenue engines that turn AI promise into durable margin.

Policy and strategy are catching up to the economics. The White House ruled out any AI bailouts, putting the onus on performance, not protection. APEC leaders endorsed a regional AI framework, while China and India pushed their own governance models. Microsoft’s $15 billion UAE infrastructure pledge and SoftBank’s joint venture with OpenAI in Japan show how national AI agendas are being underwritten by corporate balance sheets.

The real AI debate now happens in public, and social signals is our weekly readout from operators and builders. This week, Andrew Ng on value from connected data, Allie K. Miller on budgets tilting toward outcome systems, and Sam Altman on revenue run rate and the multi-year infrastructure that supports it. The takeaway is clear: capital discipline, contracted compute, and AI that earns its keep inside operations.

Here’s your Saturday guide to the signals shaping the future of AI:

Infrastructure

• OpenAI signs a seven-year $38 billion cloud deal with AWS. OpenAI will tap hundreds of thousands of NVIDIA GPUs in Amazon data centers to train and run models like ChatGPT, diversifying beyond Microsoft Azure and sharpening hyperscaler competition. Click here
• Microsoft signs a $9.7 billion AI cloud deal with IREN. The five-year pact gives Microsoft access to NVIDIA chips via IREN’s Texas campus, includes a 20 percent prepayment, and pairs with a $5.8 billion Dell procurement for delivery. Click here
• Lambda and Microsoft strike a multibillion-dollar AI infrastructure deal. The pact will deploy tens of thousands of NVIDIA GPUs, including GB300 NVL72 systems, to expand Azure capacity for training and inference. Click here
• Cipher Mining secures a $5.5B, fifteen-year AWS AI hosting lease. The deal covers 300 MW of turnkey space and power for AI, with delivery staged through 2026 and rent starting in August, locking in long-term revenue. Click here

Enterprise

• Anthropic projects $70 billion in revenue and $17 billion in cash flow for 2028. The Information reports Anthropic raised its internal “optimistic” forecasts this summer, targeting $70B in 2028 revenue, up from about $5B this year, alongside $17B in cash generation. Click here
• OpenAI reaches 1 million business customers. OpenAI says more than 1 million businesses now use its products, with customers like Amgen, Commonwealth Bank, Booking.com, Cisco, Lowe’s, Morgan Stanley, T-Mobile, Target, and Thermo Fisher, signaling adoption beyond pilots. Click here
• Cognizant rolls out Anthropic’s Claude to up to 350,000 employees. The firm will use Claude across corporate, engineering, and delivery teams, and deploy Claude Code to speed coding, testing, and documentation while helping clients move from pilots to production. Click here
• SAP launches a business-first AI model and deepens its Snowflake tie-up. At TechEd, SAP unveiled SAP-RPT-1, a table-based model that predicts outcomes like late deliveries and payment risk, and said it will train 12 million people in AI skills by 2030. Click here
• Palantir reports 121 percent growth in U.S. commercial revenue to $397 million in Q3. The surge reflects accelerating adoption of its AIP platform as enterprises pay for AI systems that drive measurable efficiency and real-world ROI. Click here

Capital Flows

• UnifyApps raises $50 million to build an enterprise OS for AI. The Series B, led by WestBridge, with ICONIQ participating, brings on Sprinklr founder Ragy Thomas as co-CEO to scale integrations that automate work across systems like Salesforce and Workday. Click here
• Armis raises $435 million in a pre-IPO funding round at a $6.1 billion valuation. The cybersecurity and AI vendor chose to stay independent amid buyout interest, citing a receptive IPO window and accelerating enterprise AI adoption to scale further. Click here
• MoEngage raises $100 million Series F led by Goldman Sachs Alternatives and A91 Partners. The customer engagement platform will use the capital to expand globally and add more AI to its Merlin suite for enterprise campaigns. Click here
• fal acquires YC-backed Remade to expand creative AI tooling. The developer-focused generative media platform is buying Remade’s AI workspace for designers to fold its canvas and workflow features into fal’s enterprise offerings and speed go-to-market. Click here

Research

• Anthropic shows early AI self-monitoring in Claude. New research finds Claude Opus 4 and 4.1 can sometimes detect injected “thoughts” before they appear in output, tell apart internal vs external inputs, and even modulate its own internal representations, though the ability is still limited and unreliable. Click here
• An AI system speeds radio signal analysis by roughly 600x. Built with NVIDIA’s Holoscan and running at the Allen Telescope Array, the new pipeline processes data faster than real time, improving accuracy and reducing false positives, lowering time and compute costs for large scientific surveys. Click here
• Enterprises pivot to smaller, efficient models to cut AI costs. Teams are choosing compact models like Anthropic’s Claude Haiku 4.5 and IBM’s Granite 4.0 Nano to match core tasks at lower cost, and Gartner expects small, task-specific models to be used three times more than general LLMs by 2027, while many firms still struggle to meet ROI goals. Click here
• AI designs functional antibodies from scratch. David Baker’s team at the University of Washington’s Institute for Protein Design used RFdiffusion to generate de novo antibodies that bind targets like influenza hemagglutinin, pointing to faster, cheaper antibody discovery for the $200 billion antibody market. Click here

Policy

• U.S. moves to block Nvidia’s sale of scaled-down B30A AI chips to China. The decision, reported by The Information, would halt planned exports and prompt a redesign effort, while China pushes state-funded data centers to use domestic chips only. Click here
• APEC leaders back a regional AI framework, China pitches a global AI body. The 21 economies adopted the Gyeongju AI Initiative to promote trusted, interoperable AI and capacity building, while President Xi proposed a World AI Cooperation Organization to shape global rules. Click here
• White House AI czar rules out an AI bailout. David Sacks said the administration will not backstop failing AI firms, citing a competitive market with multiple frontier labs, and will focus on faster permitting and power buildouts without raising household rates. Click here
• India publishes a national AI governance framework built on “do no harm.” MeitY’s guidelines stress human-centric, safe AI with lifecycle risk checks, transparency, and audit requirements for developers and deployers. Click here

Global AI Strategy

• SoftBank and OpenAI launch SB OAI Japan to deploy “Crystal intelligence.” The 50-50 joint venture will localize OpenAI’s enterprise technology for Japan, offering an integrated AI solution to modernize management and operations across industries. SoftBank will be the first adopter before the broader rollout in 2026. Click here
• EU unveils the Apply AI Strategy to speed adoption across key industries. The European Commission’s plan prioritizes deploying AI in health, energy, and manufacturing, with initial funding from Horizon Europe and Digital Europe (about €1 billion), plus skills and cross-border data initiatives aligned with the AI Act. Click here
• Microsoft pledges $15 billion for UAE AI infrastructure by 2029. The plan expands Azure data centers (about 200 MW of additional capacity) and brings in U.S.-approved NVIDIA GPUs through Microsoft’s partnership with G42, positioning the UAE as a regional AI hub. Click here
• Vanguard makes Hyderabad its new tech and AI center. The firm opened its first India hub and plans to grow to about 2,300 roles, building AI, data, and cloud tools that plug into its global operations. Click here

📱Social Signals

The most important conversations in AI are unfolding across social media, where top voices are shaping the next wave of signals and strategy. Here are some of the top social signals and their takes from the past week.

• Andrew Ng (Click here) — “The value you can now create from ‘connecting the dots’ between different pieces of data is higher than ever." Ng argues that AI agents unlock exponential ROI only when data flows freely across silos—yet SaaS vendors charge $20K+ per API connection and lock customers in, forcing enterprises to reclaim data ownership to capture cross-system outcomes and build defensible, high-margin business models.
• Allie K. Miller (Click here) — “The top average investments are from banking/finance, tech/telco, and professional services.” Miller’s Wharton 2025 AI Adoption Report data shows enterprises consistently allocate ~40% of AI budgets to tech/tools (new + existing), signaling a mature focus on outcome-enabling systems over raw infra; yet 12-15% to consultants and only 14-15% to hiring reveals a talent and execution gap that risks stalling ROI at scale.
• Sam Altman (Click here) — “I would like to clarify a few things. First, the obvious one: we do not have or want government guarantees for OpenAI data centers. We expect to end this year above $20 billion in annualized revenue run rate and grow to hundreds of billions by 2030. We are looking at commitments of about $1.4 trillion over the next 8 years.” Altman pushes back on rumors of public subsidies while revealing OpenAI's explosive growth trajectory—over $20 billion ARR this year scaling to hundreds of billions by 2030—against $1.4 trillion in private infra pledges, signaling a high-stakes bet that enterprise deals and AI-as-a-service will deliver outcomes justifying the world's largest capex without taxpayer backing.

To go deeper, subscribe to my monthly Founder Insights newsletter, where I share lessons from the frontlines of company building, perspectives on AI's future, and our industry's road ahead: https://www.linkedin.com/newsletters/founder-insights-7274531066957217793/

↓ In the comments below, share which areas in AI you want covered next week.