When 62% of enterprises are already experimenting with AI agents and 96% plan to expand their use within the year, according to Cloudera, a question emerges that no one’s answering: who orchestrates the agent swarm?
Not which model runs the agents. Not which cloud hosts them. The question Wall Street is missing: when you have 50+ autonomous agents running across IT, HR, customer service, security, and finance—who governs them?
ServiceNow just spent $11.6 billion answering that question. And the market called it “healthy but not enough.” They’re wrong.
ServiceNow isn’t buying market share in saturating categories. They’re buying the complete stack—identity, visibility, orchestration—to become the control point for enterprise agentic AI. This isn’t M&A desperation. It’s NVIDIA’s playbook applied to the workflow layer.
In Is Software Dead? Who Survives When AI Eats the Premium?, I wrote that I was “watching the orchestration and observability space more closely than the AI copilot space.” I also introduced Avenir’s framework arguing that AI value creation requires Intelligence, Context, and Agency—and that SaaS companies must evolve from “systems of record” to “systems of context” or be reduced to dumb data stores. This piece is the case study. ServiceNow is executing that evolution in real time, and they’re spending $11.6 billion to make it irreversible.
The Market Is Handing You a Gift
Let’s start with the price action, because it frames everything that follows.
ServiceNow trades at approximately $107 as of February 17, 2026—down roughly 30% year-to-date and nearly 50% from its 52-week high of $211.48. The forward P/E has compressed to approximately 43x, well below the stock’s five-year average of 68x. Software multiples broadly have compressed 33% since October 2025, according to Morgan Stanley. Wedbush’s Dan Ives calls it “the worst selloff in 25 years” and “the most disconnected trade I’ve ever seen in my career on Wall Street.” He’s calling ServiceNow and Salesforce “historic buys.”
The market’s thesis: AI agents from Anthropic, OpenAI, and others will automate enterprise workflows and render traditional software platforms obsolete. Gartner’s cautious 2026 enterprise IT spending guidance and the release of Anthropic’s Claude legal plugin in early February accelerated the panic. “SaaS is cooked” has become the prevailing narrative on fintwit and in hedge fund conversations. OpenAI’s Frontier platform, Anthropic’s expanding enterprise capabilities, and the rise of autonomous AI agents are all cited as evidence that traditional software business models are on borrowed time. The IGV software ETF has slid nearly 30% from its September highs. Goldman Sachs analysts have reportedly compared the software selloff to the decline of newspapers and tobacco companies.
The contagion is spreading well beyond software. Investors are fleeing what were until recently premium-rated, high-fee, labor-intensive service sector models—the panic has jumped from business software to tax planning, wealth management, and even commercial real estate services. CBRE’s CEO mentioned on an earnings call that AI could reduce long-term demand for office space; the stock dropped 20% in two days despite beating estimates. Insurance stocks sold off after OpenAI added a couple of basic insurance quote apps to its ChatGPT store. Over the last eight trading sessions through February 14, 115 stocks in the S&P 500 slumped 7% or more in a single day—while the index itself was roughly flat. The average stock in the S&P 500 moved approximately 10% over the past two weeks, a 99th percentile level of inter-sector dispersion versus index volatility. We’re in the phase where investors are selling first and thinking later, if at all. Nuance is being lost in the noise.
The simplest version of the bear case: every company that doesn’t touch physical reality is just expensive middleware between a problem and a solution. AI offers pure digital output at near-zero marginal cost, 24/7. If your product is information, your company is just organized thinking—and thinking can now be replicated at near-zero cost. Under this framework, software and service companies that scaled on near-zero marginal cost, strong pricing power, and minimal capital requirements face a secular mean reversion as barriers to entry erode and software itself becomes more capital-intensive. Private equity firms that loaded up on SaaS at 30x EBITDA during the 2021 peak are now staring at median multiples closer to 16x, savage write-downs, and the existential question of whether their portfolio companies can survive agentic AI disruption.
But here’s what the indiscriminate selling misses. The same macro strategists warning that asset-light software is being commoditized are simultaneously identifying the categories that will thrive in the AI era: enterprise systems-of-record where switching costs and data gravity lock users in, payments and identity and authentication networks sitting inside transaction flows, and platforms where proprietary data is continuously generated and cannot be easily replicated. As intelligence becomes cheap, proprietary information becomes more valuable—data ownership, access, and positioning within real economic flows are what matter. And implementing AI effectively in the enterprise still requires extensive “replumbing” of corporate data flows—Anthropic’s own engineers reportedly spent six months embedded at Goldman Sachs just to build autonomous systems for back-office work, because to make agents effective, they have to be integrated with non-public data silos.
Read that list of survivors again: systems-of-record, identity networks, authentication layers, platforms with proprietary data gravity, and companies positioned within real economic flows. That’s not a description of a random software company at risk of disruption. That’s a description of exactly what ServiceNow is building.
And the executives closest to the business know it. They aren’t selling. They’re buying.
CEO Bill McDermott has committed approximately $23 million of personal capital to purchasing ServiceNow shares. According to WebProNews, he committed $20 million in late January alongside a pledge to lead the company through 2030. Then on February 13, per an SEC 8-K filing, he entered an agreement to purchase an additional $3 million at prevailing market prices on February 27—the earliest date he could buy without triggering short-swing profit rules under Section 16. These are open-market purchases with after-tax personal funds, not stock grants or options exercises.
But it goes beyond McDermott. Five senior executives—including CFO Gina Mastantuono, Vice Chairman Nicholas Tzitzon, Chief People and AI Enablement Officer Jacqueline Canney, and Special Counsel Russell Elmer—simultaneously terminated their Rule 10b5-1 trading plans, cancelling all future pre-planned stock sales. In an environment where insiders have been steady sellers for years, the entire C-suite collectively decided to stop selling and the CEO started buying. That’s not a routine corporate governance move. That’s a signal.
On the Q4 earnings call in late January, McDermott told investors directly: “The worry is gone, you can give us back the market cap.” The market didn’t listen. So the executives put their own money where their conviction is.
The stock’s reaction on the day of the announcement? It dropped another ~2.5% intraday. The market is so conditioned to fear the “software is dead” narrative that it can’t process a bullish signal when one arrives. That’s when contrarian opportunities emerge.
The Agent Chaos Problem
Inside ServiceNow’s own operations, according to the company, AI agents now resolve 90% of IT requests and 89% of customer support tickets autonomously. No human in the loop. The productivity gains are real. But scale this across an enterprise with hundreds of agents, and you hit three problems simultaneously.
The first is the identity problem. Traditional identity and access management systems weren’t designed for non-human agents that spawn, execute tasks, access sensitive data, and terminate—sometimes in minutes. When an AI agent requests access to your financial systems at 3 AM, how do you verify it’s authorized? Legacy IAM treats every request as human-initiated. Agent-native identity governance doesn’t exist in most enterprises.
The second is the visibility problem. Your IT team can inventory laptops and servers. But can they inventory every IoT sensor, operational technology system, and medical device connected to your network? When an AI agent needs to orchestrate workflows across IT, OT, and IoT environments, you need real-time visibility into assets you didn’t even know existed. Gartner has estimated that approximately 40% of enterprise-connected devices are unmanaged or unknown.
The third is the orchestration problem. You’ve deployed agents from OpenAI, Anthropic, Google, Microsoft, and your own custom models. They’re running on different platforms with different APIs and different security models. When Agent A (HR onboarding) needs to hand off to Agent B (IT provisioning) which then triggers Agent C (security compliance), who manages that workflow? Who ensures the agents don’t conflict, duplicate work, or create security gaps?
Regular readers will recognize these three layers. In The Verification Gap: Who Audits the Agent Swarm?, I identified orchestration, verification, and security as distinct control challenges that “often get conflated.” I argued that orchestration was largely solved for small deployments, but verification and security remained wide open. ServiceNow appears to be building the first platform that addresses all three simultaneously—and they’re doing it through acquisitions rather than hoping the open-source community closes the gap.
Wall Street analysts saw ServiceNow’s +19.5% subscription revenue growth (constant currency) and +21% cRPO growth (constant currency) and called it steady but unexciting. They missed the signal: ServiceNow isn’t optimizing the old business. They’re building the infrastructure layer for an entirely new computing paradigm.
Four-Layer Stack Depth
NVIDIA didn’t become a $3 trillion company by selling faster chips. They won by building CUDA—the software layer that made their hardware essential. ServiceNow is executing the same strategy, but for workflows instead of compute.
Layer 1: Research (Foundational IP)
ServiceNow AI Research isn’t a marketing team writing blog posts. They’re publishing at NeurIPS. They’re collaborating with Stanford, University of Toronto, McGill, and Université de Montréal. They’re founding members of the AI Alliance. Most importantly, they’re building proprietary foundation models. The Apriel family and Now LLM represent ServiceNow’s bet that orchestration requires model ownership, not just model consumption. You can’t govern what you don’t understand at the architecture level.
This matters because orchestration isn’t prompt engineering. It’s understanding how different models reason, where they fail, and how to route tasks to the right model for the job. When ServiceNow’s AI Control Tower decides whether to route a request to Claude, GPT-5, or Gemini, that decision requires deep model understanding—the kind you only get by building models yourself.
Layer 2: Product (The Control Point)
AI Control Tower launched at Knowledge 2025, and according to management on the Q4 2025 earnings call, deal volume nearly tripled and exceeded internal expectations. That’s not incremental adoption. That’s enterprises recognizing they have an agent governance crisis.
The Control Tower provides what no other platform offers: unified visibility and governance across any AI agent, any model, any workflow. ServiceNow or third-party. OpenAI or Anthropic. Custom or commercial. The platform sees them all, monitors their behavior, enforces policies, and generates compliance reports. Think of it as mission control for your agent fleet. Real-time dashboards show which agents are running, what they’re working on, which data they’re accessing, and whether they’re delivering business outcomes. When an agent violates a security policy, the Control Tower can pause it, require human approval, or terminate it entirely.
AI Agent Fabric, launching Q3 2025, takes this further: agent-to-agent communication and collaboration. When your HR agent needs to trigger your IT provisioning agent which needs to notify your security compliance agent, Agent Fabric handles the orchestration. No custom integration code. No fragile API chains. The platform manages the workflow.
Now Assist surpassed $600 million in ACV by Q4 2025, according to CEO Bill McDermott on the earnings call, more than doubling year-over-year. It’s tracking toward $1 billion+ by year-end 2026. That figure represents the Pro-to-Pro Plus upsell and excludes Moveworks contribution and consumption-based revenue from Now Assist packs. In Q4 alone, deals greater than $1 million nearly tripled quarter-over-quarter, and the number of deals including five or more Now Assist products increased by over 10x year-over-year. The AI products are scaling faster than the core platform did.
Build Agent, powered by Anthropic’s Claude as the default model, enables developers to build and deploy agentic workflows with a targeted 50% faster implementation cycle. ServiceNow expects Build Agent usage to quadruple over the next 12 months. When your platform makes it easier to build agents and easier to govern them, you’ve created a flywheel.
Layer 3: Acquisitions (Stack Depth at Scale)
$11.6 billion in acquisitions over 12 months isn’t random. It’s surgical.
Moveworks closed in December 2025 for $2.85 billion. It brings agent orchestration capabilities and a Reasoning Engine that routes tasks across agents intelligently. Moveworks’ front-end AI assistant and enterprise search capabilities fill the user interface gap. ServiceNow gets the orchestration backend and the conversational frontend in one deal.
Veza, reportedly valued at approximately $1 billion according to SecurityWeek’s M&A tracker (financial terms were not officially disclosed), is expected to close H1 2026. It solves the identity governance problem for non-human agents. As one industry analyst noted, “Identity governance has become a real bottleneck preventing organizations from deploying AI agents at scale.” Veza’s identity and access management for AI agents—powered by its patented Access Graph that maps relationships across human, machine, and AI identities—isn’t a feature. It’s the foundation of secure agentic AI.
Armis, at $7.75 billion in cash, is expected to close H2 2026. Armis provides real-time visibility and risk prioritization across IT, operational technology, medical devices, and industrial environments. The company has surpassed $340 million in annual recurring revenue with year-over-year growth exceeding 50%, and its platform is trusted by over 35% of the Fortune 100. When your agents need to operate across the full technology surface—from cloud workloads to factory floor sensors—Armis delivers the visibility layer.
Together, these three acquisitions deliver visibility (Armis), identity (Veza), and orchestration (Moveworks). Management explicitly framed this on the Q4 earnings call. McDermott stated: “Our announced plans to acquire Veza and Armis happened in rapid succession because this assembles three critical layers for enterprises to operate securely in an agentic AI world: visibility, identity, and orchestration.”
McDermott himself validated this framework even more directly. In a message reportedly sent to a long-term shareholder who inquired about the company’s outlook, the CEO wrote: “The biggest surprise will be ServiceNow as the control tower for business reinvention. The reason we went for Veza and Armis is because the defining challenge of the next decade and beyond is to identify, orchestrate and secure the billions of AI agents. With our CMDB as the bedrock foundation of the modern enterprise, we will win.”
That language—identify, orchestrate, secure—maps precisely to the three acquisition layers. It also reveals something the earnings call didn’t emphasize: McDermott sees CMDB (Configuration Management Database) as the structural advantage. CMDB is ServiceNow’s system of record for every asset, relationship, and dependency in an enterprise. When you layer identity governance (Veza), asset visibility (Armis), and agent orchestration (Moveworks) on top of a CMDB that already maps the enterprise, you don’t just get a control tower—you get the only control tower with a complete model of the business underneath it.
This is NVIDIA’s Mellanox playbook. As I detailed in NVIDIA’s Inference Stack Depth Strategy, NVIDIA didn’t just build faster GPUs. They acquired the networking layer (Mellanox for $7 billion), the orchestration layer (Run:ai), the optimization pipeline (Deci), and inference IP (Groq licensing). You can’t separate the GPU from the network fabric from the scheduling software. They co-designed the full stack. ServiceNow is doing the same thing for workflows. You can’t separate agent orchestration from identity governance from asset visibility. They’re co-designing the enterprise AI control plane.
Layer 4: Partnerships (Model-Agnostic Moat)
In late January 2026, ServiceNow announced Claude as the default model for Build Agent as part of a multi-year partnership with Anthropic. This came days after a separate collaboration announcement with OpenAI, and alongside existing partnerships with Microsoft, Google, and NVIDIA. The strategy is clear: ServiceNow wins when it orchestrates any model, not when it locks customers into one vendor.
The Control Tower supports agents built on OpenAI, Anthropic, Google, Microsoft, or custom models. The platform is model-agnostic because the value isn’t in the model—it’s in the orchestration layer. This mirrors CUDA’s strategy. NVIDIA doesn’t care which framework you use—TensorFlow, PyTorch, JAX. CUDA supports them all. That’s why it became the standard. ServiceNow is building the CUDA for workflows: the layer that every model depends on but no model can replace.
The Anthropic partnership is particularly revealing. ServiceNow deployed Claude to 29,000 employees for sales preparation, achieving up to 95% reduction in prep time. They made Claude Code available to engineering teams. This isn’t vendor selection—it’s field testing at scale. ServiceNow is learning which models excel at which enterprise tasks, then encoding that knowledge into AI Control Tower’s routing logic. Healthcare claims authorization could drop from days to hours using Claude-powered agents on ServiceNow’s governed platform. Research analysis in life sciences accelerates while maintaining compliance. The partnership validates that the orchestration layer delivers value even with constantly improving models underneath.
The Co-Design Thesis: Workflows Edition
I’ve written extensively about co-design in the AI infrastructure stack—how NVIDIA wins because they design hardware, software, and deployment as a unified system rather than independent components. You can’t optimize the GPU without optimizing the network. You can’t optimize the network without optimizing the software stack. The boundaries dissolve.
ServiceNow is applying the same principle to enterprise workflows.
You can’t govern agents without governing identity. When an agent requests database access, the governance system needs to know: which human authorized this agent? What permissions should it inherit? How long should the access last? Traditional IAM systems don’t have answers. Agent-native identity governance (Veza) solves this.
You can’t govern identity without visibility. If you can’t see all the assets on your network—the unmanaged IoT devices, the OT systems, the shadow IT—you can’t enforce identity policies. An agent that connects to an unknown device creates a security gap that traditional tools miss. Real-time asset visibility (Armis) closes this gap.
You can’t orchestrate workflows without both. When Agent A hands off to Agent B, the orchestration layer needs to verify: does Agent B have permission to access the resources it needs? Are those resources visible and properly inventoried? Can we audit the full workflow for compliance? Orchestration (Moveworks + AI Control Tower) ties it together.
This is why the acquisitions cost $11.6 billion. ServiceNow isn’t buying features. They’re buying the pieces of a system that only works when designed together. You can’t bolt identity governance onto an orchestration platform as an afterthought. You can’t add visibility as a plugin. The architecture has to be unified from the ground up.
Wall Street analysts are valuing ServiceNow as an application software company growing at 20%. They should be valuing it as an infrastructure platform controlling a $15.5+ billion subscription revenue base (FY26 guidance) with 36% free cash flow margins that’s becoming the standard control plane for enterprise AI.
This is what I meant in Is Software Dead? when I wrote that the winners in the AI era would be companies that evolve from “systems of record” into “systems of context.” Avenir’s framework identified three components of AI value creation: Intelligence (commoditizing rapidly), Context (the battleground), and Agency (requires context as a precursor). ServiceNow is systematically capturing all three. The proprietary models (Apriel, Now LLM) give them Intelligence. The CMDB, Veza’s Access Graph, and Armis’s asset discovery give them Context—the richest enterprise context of any platform. And AI Control Tower plus Moveworks give them Agency, the ability to act autonomously across workflows. McDermott calls it “the semantic layer that makes AI ubiquitous in the enterprise.” The Avenir framework explains why that positioning works: you can’t delegate tasks to AI agents without giving them the context to execute. ServiceNow owns the context.
Why “Healthy but Not Enough” Misses the Point
FundaAI’s report called ServiceNow’s Q4 results “healthy but not enough” because cRPO beat guidance by approximately 2 points versus wider beats in prior quarters. FY26 subscription revenue guidance of 19.5-20% constant currency growth came in below the 22% whisper number bulls had priced in. The bears see subscription gross margin guidance of approximately 82% from AI product mix and call it a warning sign. They see $11.6 billion in M&A and worry about core saturation.
They’re analyzing the wrong business.
ServiceNow is deliberately investing in AI infrastructure—the research team, the foundation models, the Control Tower platform. These are investments in future platform economics, not signs of margin pressure. The M&A isn’t desperation—it’s stack depth. NVIDIA’s Mellanox acquisition initially pressured margins too. Now it’s the networking layer that makes NVIDIA’s data center platform unassailable. ServiceNow is making the same bet: spend heavily now to own the orchestration layer before competitors realize it’s the choke point.
FY26 guidance of ~20% subscription revenue growth includes only a 1-point contribution from Moveworks (which closed in December 2025). It largely excludes Veza (closing H1 2026) and Armis (closing H2 2026, expected to contribute approximately 1 point of subscription revenue in 2026). The organic business is steady while ServiceNow layers in $11.6 billion of strategic assets. That’s not deceleration—that’s digestion.
Now Assist hit $600+ million ACV, more than doubling year-over-year. AI Control Tower deal volume nearly tripled. Build Agent usage is projected to quadruple in 12 months. Deals including five or more Now Assist products increased over 10x year-over-year. These aren’t incremental product launches. They’re platform inflection points.
The $5 billion share repurchase authorization (including a $2 billion ASR) signals management’s confidence at the corporate level. McDermott’s $23 million in personal stock purchases signals it at the individual level. And the entire C-suite cancelling their selling programs signals it collectively. They’re simultaneously deploying $11.6 billion on strategic M&A, returning $5 billion to shareholders, and putting their own net worth on the line. Full-year 2025 free cash flow was $4.6 billion, up 34% year-over-year, with operating margin at 31%. The company ended 2025 with over $10 billion in cash and investments. That’s the behavior of a company—and a leadership team—that sees a generational dislocation between price and value.
The Investment Framework
The question isn’t whether agentic AI will transform enterprise workflows. With 96% of enterprises planning to expand agent deployments, that’s settled. The question is: who controls the orchestration layer?
So What?
Wall Street is valuing ServiceNow as a 20% growth application software company with margin pressure. That’s the wrong frame.
ServiceNow is building the orchestration layer for enterprise AI—the platform that governs identity, ensures visibility, and coordinates workflows across any agent, any model, any environment. They’ve spent $11.6 billion acquiring the complete stack and they’re generating $1 billion in new AI product revenue within 18 months of launch.
When analysts say “healthy but not enough,” they’re measuring the old business. The new business—the control plane for agentic AI—is just beginning to scale. And unlike the model layer (where OpenAI, Anthropic, and Google compete) or the infrastructure layer (where cloud providers compete), the orchestration layer has natural winner-take-most economics. Once enterprises standardize on ServiceNow for agent governance, switching costs become prohibitive. You can swap models underneath (Claude to GPT-5). You can change cloud providers (AWS to Azure). But migrating your entire agent orchestration, identity governance, and asset visibility infrastructure? That’s ripping out the nervous system.
And there’s the CMDB angle that even the bulls haven’t fully priced in. The macro thesis currently crushing software stocks actually contains the answer to why ServiceNow survives: as intelligence commoditizes, value migrates to whoever controls proprietary data that is continuously generated and cannot be replicated. ServiceNow’s Configuration Management Database is exactly that—it’s the system of record for enterprise assets, relationships, and dependencies at thousands of organizations, and it gets richer with every ticket, every workflow, every asset change. When Veza’s Access Graph maps identity permissions, it maps them against the CMDB. When Armis discovers unknown assets, it populates the CMDB. When AI Control Tower governs agents, it governs them within the context of the CMDB. The CMDB isn’t just a database—it’s the enterprise’s digital twin. And ServiceNow is the only company building an agent orchestration layer on top of it.
This connects directly to what Jensen Huang declared at CES 2026: “Context is the new bottleneck.” At the silicon level, that meant rearchitecting storage hierarchies to feed context windows. At the enterprise level, it means something analogous—the companies that own the richest contextual data about how businesses actually operate will control the agentic future. ServiceNow’s CMDB is enterprise context. It maps every asset, every relationship, every dependency. When you add Armis (what’s connected), Veza (who can access what), and Moveworks (what’s being worked on), you get the most complete enterprise context graph in existence. That’s not a software product. That’s the substrate.
NVIDIA didn’t become a $4.5 trillion company overnight. They spent two decades building CUDA before AI made it essential. ServiceNow is in year two of building the CUDA for workflows. The market will figure this out. The only question is whether you recognize the pattern before Wall Street does.
With the stock down 50% from its highs, the CEO buying $23 million in personal shares, the entire C-suite cancelling their selling programs, and Wedbush calling it “the most disconnected trade” in 25 years of covering software—the setup is as clear as it gets. Everyone says SaaS is cooked. What they mean is that commodity SaaS—the GPT wrappers, the thin AI features, the horizontal copilots—is cooked. They’re right about that. But the orchestration layer? The platform that governs identity, visibility, and coordination for every AI agent an enterprise deploys? That’s not being disrupted by AI. It’s being created by AI. ServiceNow isn’t the victim of the agentic revolution. It’s the infrastructure.
Resources
ServiceNow Q4 and Full-Year 2025 Financial Results — ServiceNow Investor Relations
ServiceNow and Anthropic Partnership — ServiceNow Newsroom
ServiceNow to Acquire Armis for $7.75B — ServiceNow Newsroom
ServiceNow to Acquire Veza — ServiceNow Newsroom
ServiceNow Completes Moveworks Acquisition — CX Today
ServiceNow Chooses Claude to Power Customer Apps — Anthropic
From ‘AI Platform’ to ‘AI Control Tower’ — Diginomica
ServiceNow Q4 2025 Earnings Call Transcript — Yahoo Finance
ServiceNow 8-K: Executive Trading Plan Terminations and CEO Share Purchase Agreement — SEC Filing, February 17, 2026
ServiceNow Executives Cancel Stock Sales to Boost Investor Faith — Bloomberg
ServiceNow’s CEO Makes Bold $20 Million Stock Purchase — WebProNews
Dan Ives Calls AI-Driven Software Selloff ‘Most Disconnected Trade’ — Benzinga
Related BEP Research
Is Software Dead? Who Survives When AI Eats the Premium? — The Avenir framework: Intelligence, Context, Agency
NVIDIA’s Inference Stack Depth Strategy — Part 2 of the Co-Design Series: the M&A playbook ServiceNow is replicating
The Verification Gap: Who Audits the Agent Swarm? — Part 3 of the Co-Design Series: the orchestration, verification, and security layers ServiceNow is now building
The Memory Wall: Why Groq and Jamba Had to Find Each Other — Part 1 of the Co-Design Series: why memory—and context—is the real bottleneck
NVIDIA CES 2026: Six Chips, One Platform, and the Extreme Codesign Era — Jensen’s “context is the new bottleneck” declaration
What’s Changing
On March 9th, BEP Research moves to paid subscriptions — timed deliberately to NVIDIA’s GTC conference the following week, where I’ll be on the ground. Paid subscribers get exclusive real-time GTC coverage: in-depth analysis of every major announcement, on-the-ground interviews, and breakdowns of what roadmap changes mean for your portfolio. Not recaps. Real analysis, in real time.
After March 9th, the deep work — Co-Design Thesis series, investment frameworks, earnings breakdowns, supply chain deep dives, and all GTC coverage — moves behind the paywall. Free subscribers will still receive post summaries and occasional public pieces.
Early Bird: ~$350/year (13% off) — available now through March 9th only.
→ Lock in the Early Bird rate here
After March 9th, standard pricing applies: $400/year | $45/month | Founding Member: $450/year (permanent recognition + priority access to the institutional tier).
This is premium-priced on purpose. Institutional-quality research without sell-side conflicts, without advertising, without surface-level takes. If one analysis helps you size a position correctly or avoid a bad trade, the subscription pays for itself many times over.
Institutional Access: I’m developing a tier for teams and funds with multi-seat licenses, direct analyst calls, and priority Q&A. If that’s relevant to your team, reply to this email.
The early bird window closes March 9th. See you at GTC.
About the Author
Ben Pouladian is a Los Angeles-based tech investor and entrepreneur focused on AI infrastructure, semiconductors, and the power systems enabling the next generation of compute. He was co-founder of Deco Lighting (2005–2019), where he helped build one of the leading commercial LED lighting manufacturers in North America. Ben holds an electrical engineering degree from UC San Diego.
He currently serves as Chairman of the Leadership Board at Terasaki Institute for Biomedical Innovation and is a YPO member. His investment research focuses on AI datacenter infrastructure, GPU computing, and the semiconductor supply chain. Long-term NVIDIA investor since 2016.
Follow on Twitter/X: @benitoz | More at benpouladian.com
Disclosure: The author holds positions in NVIDIA and related semiconductor investments. The author does hold a position in ServiceNow. This is not investment advice.
I’ve never been an investor in NOW; however, it’s always been of interest. I hold a small position in PATH and, putting aside differences of approach, I’ll just say you’ve outlined exactly what my thoughts are around the agentification of enterprises.
The contextual visualisation of business processes and then identifying how and what can be agentified in a secure and compliant way is key. It seems that NOW understands that and, rather than solely developing in-house (despite having significant internal build capability), the smart choice in certain instances is to acquire in order to gain that knowledge and expertise more rapidly and defensibly.
An example I often tout is that the exec or higher management within a company has probably used ChatGPT (or Claude/Gemini) and realised that “AI,” in some capacity, can make their organisation more efficient and productive. However, they won’t fundamentally understand the intricate mechanics that weave together orchestration (systems integration), compliance, security, and outcome.
So, through a potentially arduous and/or unsuccessful process of agentic transformation, they inevitably realise they require the structured, secure, and enterprise-grade framework that NOW or PATH provide, not experimentation, but infrastructure.
Great read, very interesting 👍