The Agentic Enterprise AK · Morning Edition · 7 min read | Wednesday, August 19, 2026 Enterprise AI finally has an income statement. Palantir posted 93% revenue growth and raised guidance the same week adoption data showed roughly 80% of new enterprise apps now ship with an AI agent. The ROI argument is over. The question is who actually collects it. For two years the honest CIO question about AI was whether it would ever pay for itself. This week produced the cleanest answer yet, and it is yes, on an income statement, at scale, in dollars a board can read. Palantir's commercial book is the proof. But the same week's adoption data complicates the celebration: agents are now embedded almost everywhere and scaled almost nowhere, and the returns are concentrating in a small group of companies that solved ownership, data, and governance before they solved for ambition. The ROI question is settled. The distribution of that ROI is the new contest. | Enterprise AI just posted the quarter that ends the ROI argument. | P | alantir reported $1.935 billion in second-quarter revenue, up 93% year over year, with US commercial revenue up 149% and a Rule of 40 score of 155%, then raised full-year guidance to 82% growth. In the same window, adoption data showed roughly 80% of enterprise applications shipped or updated this year now embed at least one AI agent. One number proves AI can print revenue; the other proves it is everywhere. Together they close a debate that has run since 2023. |
Strip the stock noise and Palantir's result is the cleanest evidence yet that AI spend has crossed from experiment to line item. This is not a demo or a pilot budget. It is $3.37 billion in total contract value for the quarter, a record $2.13 billion of it US commercial, booked against real deployments with real outcomes attached. When a vendor packages AI as a business result rather than access to a model, enterprises are now signing large, multi-year commitments for it. Here is the part the headline growth rate hides. The revenue is real, but it is narrow. It concentrates in the vendors and buyers who treat deployment as an operating discipline, not a technology purchase. Ubiquitous embedding is not the same as captured value, and the gap between the two is where most AI budgets will be won or quietly written off over the next year. The ROI question is settled. The distribution of that ROI is the new contest. |
The Spearhead Take Stop funding pilots that prove feasibility. Fund the two or three use cases you can measure on a P&L, put a single owner on each, and buy from vendors who sell outcomes rather than endpoints. The companies booking AI revenue this quarter share one trait, and it is not model choice. It is single-threaded ownership tied to a number a CFO already tracks. |
| The Obvious & The Overlooked Three reads the market has made. Four it has not. The Obvious Palantir proved enterprise AI can print money. A 93% quarter with raised guidance is the clearest revenue signal the category has produced. Business WireAgents are now default, not novel. Roughly 80% of enterprise apps shipped this year embed at least one agent. Digital AppliedWall Street is underwriting the buildout. Nvidia lined up more than $500 billion of third-party capital for AI infrastructure. CNBC | The Overlooked The revenue is concentrated, not shared. Only about 3% of companies have scaled agents across departments, so a minority is capturing most of the returns. Agentic AI InstituteThe constraint is shifting to FinOps. Snowflake shipped a Cortex AI Gateway whose job is to stop agents from running up runaway cloud bills. VentureBeatGoverning agents is now its own budget line. Obsidian's $85 million round rests on 60 of the Fortune 500 paying to control what their agents can touch. Unite.AIAdoption tracks measurability. Financial services leads at 95% deeper adoption because it can put a number on the outcome; fuzzier-ROI sectors lag. PYMNTS |
| Moving Pieces Five developments worth a CIO's attention. InfrastructureWall Street agreed to finance the buildout, and Nvidia will backstop a quarter of it Nvidia signed memoranda with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish platforms designed to mobilize more than $500 billion of third-party capital for AI infrastructure, with Nvidia potentially backstopping up to $125 billion. The move reclassifies compute as an asset class, financed like real estate against long-dated debt. For a buyer, financed capacity should ease scarcity, but it is only as durable as the demand assumptions behind the loans. The capacity your 2027 roadmap assumes now depends on who holds the paper. ProductRyanair puts Gemini on 35,000 desks and keeps AWS as a spare Europe's largest airline signed a five-year deal to deploy Gemini Enterprise and Google Workspace to 35,000 employees, using Google DeepMind models for crew logistics, fleet operations, and maintenance planning. The detail worth copying is architectural: Ryanair extended AWS weeks earlier and is running Google as a deliberate second hyperscaler so critical services can fail over during an outage. Agentic workloads are pushing even cost-obsessed operators toward multi-cloud, because a single-vendor agent stack is now a single point of failure. InfrastructureIBM enters the neocloud business on Together AI's dime IBM signed a $240 million multi-year deal to build a dedicated Nvidia HGX B300 inference cluster on IBM Cloud for Together AI, roughly 2,000 Blackwell chips live in the US by Q1 2027. The read is that IBM, long a managed-services incumbent, is now renting raw accelerated compute to a neocloud rival of the hyperscalers it partners with. When Big Blue starts selling GPU capacity by the cluster, the line between enterprise IT vendor and compute landlord has effectively dissolved. Governance / SecurityObsidian raises $85 million to police the agents everyone is deploying Obsidian Security closed an $85 million Series D at a $1.1 billion valuation to expand runtime governance for AI agents inside SaaS, counting 60 of the Fortune 500 as customers. Notably, it is extending native controls to Anthropic's Claude Code and Cowork, letting security teams restrict agent access to production data and block unsanctioned tool use. The category tell is that agent security is now funded like a platform, not a feature. If you are deploying agents faster than you can govern them, the market has already priced that gap. PolicyThe EU can now fine model makers, and the meter is running As of August 2, the EU AI Act's enforcement powers over general-purpose AI providers went live. The Commission's AI Office can demand documentation, run technical evaluations, force mitigation, pull a model from the market, and levy fines up to 3% of global turnover or 15 million euros. Failing to respond adequately is independently sanctionable at the same ceiling. High-risk deadlines slipped to 2027 and 2028, but for any enterprise standardizing on a frontier model in Europe, the vendor's regulatory posture just became a procurement variable. | On the Radar Nine signals, sharpened. | Deals | Higgsfield raised $400 million. The generative-media startup's Series B was led by DST Global and Goldman Sachs Alternatives, a sign investors still see room beneath the model labs. Crunchbase News | | Compute | Groq raised $350 million at a $3.5 billion valuation. The inference-chip challenger keeps drawing capital as buyers look for alternatives to paying Nvidia margins on every token. Crunchbase News | | Product | Lovable raised $400 million. Menlo led the Series C for the vibe-coding platform, extending the bet that natural-language app building reaches non-developers inside the enterprise. Crunchbase News | | Deals | Wispr Flow raised $280 million. Menlo backed the voice-dictation startup at Series B, wagering that speech becomes the default input for agentic workflows. Crunchbase News | | Deployment | Cognizant stood up a dedicated EMEA AI unit. The services firm is packaging agentic delivery into Foundation, Accelerate, and Transform tiers, a tell that integrators are productizing the pilot-to-production climb. Orevia News | | Research | Anthropic raised its own misalignment rating from very low to low. Its August risk report cited rising uncertainty and disclosed an unreleased internal model it says it has no plans to ship. SiliconANGLE | | Workforce | GitLab cut about 350 roles, 14% of staff, to fund AI infrastructure. Leadership framed the reduction as freeing capital for agentic development rather than a demand problem. TechCrunch | | Product | Databricks shipped Genie Code and Snowflake previewed SnowWork. The data platforms are racing to add autonomous execution layers that build pipelines and run workflows without a human in the loop. SiliconANGLE | | Deals | Google is reported to be acquiring Spirit Airlines' data assets. If confirmed, it underscores how proprietary operational data has become the scarce input model access no longer provides. Stratechery |
| Quick Hits Ten more, worth knowing. | Atoms raised $1.7 billion led by Andreessen Horowitz. Travis Kalanick's physical-AI venture was the week's largest round. Crunchbase News | | Meshy AI raised $400 million at a $1.5 billion valuation. The 3D generative-model maker drew IDG, Monolith, and Matrix. Crunchbase News | | Augustus raised $180 million at a $1 billion valuation. Tiger Global led the round for the AI-driven cross-border banking startup. Tech Startups | | Gravis Robotics raised $200 million. Autonomous construction machinery joined the week's physical-AI theme. Crunchbase News | | Graas acquired Trustana. The Singapore ecommerce-AI firm is consolidating retail agent capabilities. blog.mean.ceo | | Salesforce added MCP and Agent2Agent support to MuleSoft. Agent orchestration is being wired into the integration layer. SiliconANGLE | | Obsidian's total funding passed $200 million. The agent-security raise pushes cumulative capital over the mark. Unite.AI | | Together AI now serves about 400 trillion tokens a month. The inference volume underlines how much enterprise workload has moved to specialized providers. The Register | | OpenAI booked $6.7 billion in Q2 sales. Revenue rose from $5.7 billion, but its operating margin slid further into the red. SiliconANGLE | | Fireworks AI closed a $1.505 billion Series D at a $17.5 billion valuation. Inference infrastructure keeps commanding premium marks. blog.mean.ceo |
| The Number 95% Of financial-services firms report deeper AI adoption The highest of any sector, ahead of healthcare at 84% and media at 81%. Adoption is tracking measurability. The sectors scaling fastest are the ones that can put a number on the outcome, which is why regulated, metrics-driven functions are pulling ahead while businesses with fuzzier ROI lag. If you cannot name the metric your AI moves, you are probably in the trailing group. | Counter-Signal GovernanceThe revenue is real. For most companies, so is the stalled pilot. The same adoption data that shows agents nearly everywhere also shows them scaled almost nowhere. Only about 3% of companies have pushed agentic AI across multiple departments, 62% are still experimenting, and just 21% report a mature governance model for autonomous agents. Data quality is the most-cited blocker, named by 52%. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027 on runaway cost, unclear value, or weak controls. Read against Palantir's quarter, the message is not that enterprise AI does not pay. It is that it pays a minority. The returns are concentrating in the companies that solved ownership, data, and governance before they scaled ambition, and the buildout of financed compute does nothing to close that gap. The takeaway is sequencing, not skepticism. Fix the plumbing and put a single owner on a measurable outcome before you scale, or you become the cancellation statistic rather than the growth one. | From the Field There is a version of this week's news that reads as pure vindication. Enterprise AI booked a blockbuster quarter, agents are in almost every new application, and the money that funds the whole thing is now flowing from the deepest pools on Wall Street. For anyone who spent two years arguing the technology would eventually earn its keep, the receipts finally arrived. We would offer a quieter reading. The gap between the companies posting 93% growth and the ones stuck at 3% scaled is not a gap in models or budgets. It is a gap in discipline. The winners picked a small number of outcomes they could measure, gave each a single owner, fixed the unglamorous data problems first, and refused to confuse a pilot that works with a program that pays. None of that is a breakthrough. It is operating hygiene, applied earlier than everyone else. So take the good news for what it is, then act on the uncomfortable part. The ROI question is answered. Whether you are in the group that collects it is still, mostly, up to you. Let's get to production,
AK | | The Agentic Enterprise Know more about AI than 95% of your peers. By 7 AM. A daily AI intelligence briefing for enterprise leaders, published by Spearhead. We build AI systems that work. Strategy. Engineering. Production. Outcomes. © 2026 Spearhead. All rights reserved. |
|
