Bruce Broussard, the interim CEO of HP, Welcome the audience to HP Imagine 2026
During the analyst Q&A at HP Imagine in New York, I had a question for HP's leadership team.
"Chatbots are becoming commoditized. The models behind them are too. And now we are watching the AI market pivot toward the next defensible layer: on-device models, local compute, and hardware partnerships. You are already there. But the same AI players racing to own the software layer have now set their eyes on the device layer too, and hardware vendors are waking up to the same opportunity. So the question I have to ask is: how long is the device as a moat actually defensible, and where does HP's uniqueness come from as everyone else races toward the same endpoint?"
The room went quiet for a moment. Then HP's executives answered.
I will get to their answer. But the more interesting question, the one I have been turning over since leaving New York, is whether HP asked themselves that question before I did, and whether their answer at HP Imagine 2026 is the beginning of something genuinely defensible, or a well-positioned opening move in a race that is just getting started.
The AI market has a device problem.
Everyone wants to own the endpoint. The companies building the most capable models are now racing to capture the desktop, the operating system, the local compute layer. The moves are coming fast and they are coming from every direction — software players, AI labs, hardware partnerships, open source agents going viral overnight. The logic is the same in every case: the model alone is not enough. To matter, AI needs to be where work actually happens.
These are not coincidental product decisions. They are a coordinated recognition that the cloud alone is not a sufficient foundation for durable AI advantage. The endpoint matters.
HP has been where work happens for decades.
At HP Imagine in New York this week, HP introduced HP IQ, an intelligence layer built directly into its device ecosystem, running a 20 billion parameter model locally, with spatial awareness, on-device memory, and enterprise governance built in from the start. It is early. It is not a finished platform. But it is the most strategically coherent answer I have seen to a question the broader AI market is still trying to figure out: where does durable advantage live when the models themselves become infrastructure?
My answer, informed by attending the last two HP Imagine events and a full day of direct conversations with HP's executive and product leadership team, is the same one HP is betting on.
It is the device. And HP is already there.
Something is shifting in how the AI industry thinks about advantage.
For the past three years the competition has been about models. Who has the best one, who can ship the next one fastest, who can price it most aggressively. That race is not over but it is changing character. The models are getting better across the board. The gaps between them are narrowing. Inference costs are collapsing. What was a differentiator eighteen months ago is becoming infrastructure today.
When the model layer commoditizes, the real competition shifts to two places. The application and experience layer, where most of the entrepreneurial energy in AI is currently focused. And the device layer, the endpoint, the place where data is actually created, where work actually happens, and where the relationship between a person and their technology is most persistent.
Those two layers are now converging. And here is why that matters for HP. When AI inference moves to the device, the application and experience layer moves with it. The context, the memory, the workflows become locally contained, tied to the physical layer rather than floating in a cloud that any competitor can also access. The application experience becomes inseparable from the device it runs on. That is a fundamentally different kind of advantage than anything a software-only AI player can create.
Every major software player is now trying to get their application layer embedded inside the device layer for exactly this reason. Anthropic has extended Claude's computer control. Cowork, to both Windows and macOS, making agentic AI accessible to everyday knowledge workers who have never written a line of code. OpenAI is collapsing ChatGPT, Codex, and its browser into a single desktop superapp, and separately acquired Jony Ive's hardware startup io for $6.4 billion to build a new generation of screenless AI devices expected later this year. If that device concept sounds familiar, it should. More on that shortly. Meta acquired Manus for over $2 billion to get into the agentic execution layer entirely. And just weeks ago at NVIDIA GTC, Jensen Huang put a spotlight on OpenClaw, the open source agent that went viral on GitHub, then answered it with NemoClaw, NVIDIA's enterprise framework for deploying local agents securely.
Every one of those moves is a software player trying to plant a flag on hardware they do not own. The endpoint is the prize. And that is exactly where HP already lives.
Prakash Arunkundrum sharing insights during HP IMagine 2026
Prakash Arunkundrum, HP's Chief Strategy and Transformation Officer (and someone I know from his previous role leading Logitech's business division), framed it plainly at HP Imagine. "AI has a gravity problem," he said. "Data is created on factory floors, in design studios, and inside mission-critical operations. Flowing between the cloud has latency, cost, and risk."
This is a sound observation about how AI economics are structurally evolving. And it points directly at HP's position. While software players are spending considerable energy trying to get onto the device, HP already ships more than half a billion of them.
The race to the endpoint is on. HP did not have to enter it. They just had to recognize they were already standing at the finish line.
If you attended HP Imagine and paid close attention, you noticed something missing. Nowhere in the main stage presentation, the press releases, or the product demonstrations was the name Humane mentioned. Not once.
It appears this was a deliberate choice worth discussing more.
In February 2025, HP acquired the assets of Humane Inc. for $116 million. Humane had raised $230 million in venture capital, was once valued at $850 million, and had sought a sale price closer to $1 billion. The AI Pin was scathingly reviewed upon launch and was bricked entirely when the acquisition closed. By most accounts, a difficult chapter.
But that is not the whole story.
What HP acquired was a correct insight attached to the wrong vehicle. Humane's founders, Imran Chaudhri and Bethany Bongiorno, are former Apple veterans who spent years thinking about what ambient, contextual, always-available AI could look like. The AI Pin tried to replace the smartphone. That was the wrong bet. People were not looking for a new device to carry. They were looking for the devices they already carry to get smarter.
It is worth pausing on that insight for a moment. OpenAI recently spent $6.4 billion to acquire Jony Ive's hardware startup, io, and build a new generation of screenless, ambient AI devices expected later this year. And their concept sounds familiar. Screenless. Contextually aware. Designed to work alongside your existing devices. HP paid $116 million for the team that already ran that experiment, learned from it, and pivoted. The lesson those founders brought with them may turn out to be worth considerably more than the acquisition price.
HP gave them the right vehicle.
I sat down with Imran and Bethany at HP Imagine and immediately found both to be charming, personable, and authentic in conversation. What came through clearly was that the pivot is not cosmetic. Imran put it directly: "It is AI plus the software and the hardware working together. Some of the other companies in this space do not have all of those components. HP does."
Imran Chaudri introducing HP IQ at HP Imagine 2026
The acquisition brought Humane's CosmOS platform, more than 300 patents, and the majority of the engineering team. Now operating out of their San Francisco lab, within three months of closing, they moved from isolation to working across every HP division-"the tip of the spear," as Bethany described it. And one year in, HP IQ is their first public output.
When I raised the Humane IP directly with Imran, I pointed to three capability buckets I had been watching since their early days. Spatial awareness, what the device perceives in the environment around it. Contextual inward sensing, what the device understands about the person in front of it. And non-UI interaction, the ability to act on intent without a button press. Imran confirmed all three are on the roadmap.
That is the part of the HP IQ story that did not make the main stage. And it is the part that matters most for the device as a moat argument.
Understanding HP IQ requires a small but important reframe. It is not a chatbot. It is not a copilot bolted onto a device. It is an intelligence layer built into the device itself, running a 20 billion parameter model locally, with no cloud dependency required for most workflows.
That distinction matters more than it might seem.
Benefits of Local AI - Image provided by HP
The knowledge worker is the initial target, and that choice is deliberate. Humane's original vision was consumer-facing. HP IQ is designed for the person who shows up to work every morning, opens an assigned device, and spends their day moving between documents, meetings, and decisions. That population is large, predictable, and sits inside enterprise environments where IT governance, security, and data sovereignty are non-negotiable requirements. It also generates enormous amounts of contextual data every day, the kind that, if captured and organized locally, becomes the foundation of something genuinely useful.
That foundation is what HP is calling the knowledge graph. The longer HP IQ operates on your device, the more context it accumulates. Your meetings, your documents, your interactions, your decisions. None of it leaves the device unless you choose to push it to a protected cloud layer. The memory builds a picture of how you work that gets more valuable over time. That is the stickiness mechanism. A cloud-based chatbot with no hardware relationship cannot replicate it.
HP IQ's universal multi-modal interface, Visor
At launch, HP IQ does several things well. It analyzes documents locally, drafts communications drawing from documents and notes you have been working on, captures and summarizes meetings without sending audio to an external server, and shares files between nearby HP devices through NearSense without cables, pairing, or setup. These are not revolutionary capabilities in isolation. What is different is where they run and what they know about you.
The roadmap is where the story gets more interesting. Three capability areas from the Humane acquisition are working their way into the product. Spatial awareness, meaning the device understands what is happening in the physical environment around it, where NearSense lives today and where the group collaboration portfolio extends naturally tomorrow. Contextual inward sensing, meaning the device builds understanding of the person in front of it through attention, presence, and interaction patterns, the path toward the digital twin use case. And non-UI interaction, meaning the room or device responds to intent without a button press or voice command. Walk to the whiteboard, the room activates. Walk into the conference room, the meeting joins. No friction, no setup.
HP IQ at work in a room environment
Greg Baribault, who leads HP's hybrid collaboration portfolio, described this vision as invisible technology. The room should just work on your behalf. That experience was not demonstrated on stage at HP Imagine. It was described in a private conversation. The gap between what HP showed publicly and what the roadmap contains is itself an analytical data point worth holding.
HP IQ is also built for the enterprise governance reality that consumer AI tools ignore. #WXP, HP's workforce experience platform, sits above HP IQ as the IT control plane. Security policies, feature controls, and fleet management all flow through WXP. The personal and corporate identity separation, two separate knowledge libraries and two separate permission sets, is actively being designed into the platform.
The architecture, in short, is the moat argument made concrete. Local compute. Persistent context. Enterprise governance. Spatial awareness expanding into group spaces. None of those four things in isolation is defensible. All four together, running on a fleet of devices that HP already owns the relationship with, is a different story.
HP IQ is a strong strategic bet. It is not a finished platform. Those two things are both true and worth holding at the same time.
Imran said it directly. HP IQ at launch is a great starting point. The team deliberately ships slightly ahead of where customers are, so that by the time users begin interrogating the capability, the next update is already ready. That is a sound product philosophy. It is also an honest acknowledgment that what was shown at HP Imagine is version one of something that needs to be considerably more capable to fulfill the moat argument.
Walk into a conference room, your device detects your presence, your calendar context is pushed to the room, the meeting joins with one click, and the room controls appear on your personal screen. No button press, no setup.
The risk, plainly stated, is shipping the org chart. If HP IQ feels like a personal productivity tool with no visible connection to how IT manages the fleet, or if the collaboration module in WXP does not look and function like the rest of the platform, the unified intelligence story falls apart at the point of experience. The branding is consistent. The integration is the work still ahead.
There is also a broader organizational question above the product work. HP has seen meaningful executive turnover in a short period. The Humane team's integration has gone well, with board-level engagement from the start and strong support from Bruce Broussard, the interim CEO. The talent is clearly present. Whether the organizational structure around that talent is stable enough to execute at the speed the strategy requires is something only time will answer.
None of this undermines the thesis. It contextualizes it. The device-as-a-moat argument is directionally correct. HP's structural position is genuinely advantaged. The question has always been about execution, and HP's history on that front is mixed enough that the honest analyst holds the conviction and the caution simultaneously.
Two years ago, I sat down with HP's leadership team at HP Imagine and came away with a clear sense that HP knew what it wanted to build but was still assembling the pieces. The strategic intent was visible. The execution path was not yet.
This year feels different.
The pieces are not all in place. But for the first time there is a coherent answer to a question the broader AI market has been circling without resolution. Where does durable advantage live when the models themselves become infrastructure? The software players are converging on the same answer HP has been sitting on for decades. It is the device. The endpoint. The place where data is created, where work actually happens, and where the relationship between a person and their technology is most intimate and most persistent.
HP's structural position is real. Half a billion devices already deployed. A trusted enterprise brand with IT relationships built over decades. A governance and management platform in WXP that is still early in its market journey but already showing meaningful customer traction, and carrying the ambition to move beyond device management into a true workforce experience and insights platform. And now, embedded in all of it, an intelligence layer built by a team that spent years thinking about exactly this problem before they had the hardware ecosystem to solve it properly.
The Humane acquisition looks different in hindsight than it did at the time. At $116 million HP did not buy a failed startup. They bought a thesis, a team, and a roadmap that is now the most strategically interesting thing happening inside one of the largest device companies in the world. Imran and Bethany are not caretakers of acquired IP. They are building something, with the full weight of HP's hardware relationships, enterprise trust, and device footprint behind them. That combination did not exist at Humane. It exists now.
Here's my short list of what I will be watching from HP as they work to close the gap between strategy and experience.
If those things come together, the device-as-a-moat thesis does not just describe HP's position. It describes a durable competitive advantage that the rest of the AI market will spend years trying to replicate from the outside.
The question the market has been asking may have an answer. It is not a better model. It is not a smarter chatbot. It is the device. And HP, for all the execution work still ahead, is the company most structurally positioned to prove it.
That is a story worth following. Let's check back here later in 2026 and see where HP is in this journey.
Disclosure: Travel and hospitality for the HP Imagine were provided by HP. Analysis and opinions are independent.
Position your event or publication with perspectives that resonate across the industry. Connect with us at the Collab Collective to learn how our experts can support your media coverage or speaking lineup.