IS AMBIENT AI THE WINNING PLAY?

Context Is King – And OpenAI Looks Exposed

In our Context is King series, we have been looking at why intelligence is now table stakes. It appears certain that the decisive factor in the AI Wars won't be which Big Tech bro builds the smartest model. It will be which becomes - or can be the most influential part of - a platform that can access, orchestrate and govern context. By which we mean all the data, documents, conversations and activities that AI systems currently can't see. Your emails. Your documents. Your Teams messages. Your calendar. Your company's institutional memory, scattered across a dozen cloud-based apps.

We found good reason for believing both Microsoft and Google are well positioned in this struggle. Redmond appears to be betting that its OS context fortress – plus the 365 productivity apps and cloud storage - is an unassailable position. Google, meanwhile, seeks to own the chips, operate the data centres, train the most competent model and continue to run the browser - all tied together with its productivity suite.

And so, as we come to write the OpenAI chapter, the landscape appears a little gloomy for Sam Altman. OpenAI currently seems to be a company that peaked early. Recent conversations at Brightbeam have included statements such as: ‘I’m not sure OpenAI has a place at the top table, longer-term.’ Sentiments which were unthinkable even a year ago. But now? Well, they’re not unreasonable.

Performance benchmarks that once felt like home turf for OpenAI have other logos at the top. And, more importantly, it is difficult to see how it could win the context war now all other forces are mobilised against it.

Google is pouring models into search results, inboxes, documents, Android and Chrome. Microsoft is wrapping intelligence around its operating system, inside 365 and across OneDrive. Anthropic has become the stable choice for organisations that want a careful intelligence in their workflows, especially ones that produce code.

Against that backdrop, OpenAI does not appear to own any obvious layer. That is not to dismiss its almost 800 million weekly users. Which is a formidable market position. It also has a name that functions as shorthand for all chatbots. A brand dominance any marketer would kill for.

But consumer attention can – and often does – switch. And OpenAI has much less direct control over the places we live and work: the OS, the browser and the productivity suite. And there are signals which could be read as the institutional realisation that context is king. Along with the understanding this is not good news.

Economic headwinds and public strain

Sam Altman has been blunt about the pressure points. A leaked memo revealed the OpenAI CEO told employees that he is bracing for ‘temporary economic headwinds’ and that the next few months could feel ‘rough’ as investor sentiment cools.

And when pressed on a Halloween podcast by host Brad Gerstner - about the chasm between the company's reported $13 billion revenue and its trillion-dollar-plus infrastructure commitments - Altman pushed back sharply thus: 'If you want to sell your shares, I'll find you a buyer'.

Microsoft's Satya Nadella, a fellow guest, laughed. And Altman later softened. But the defensiveness was unmistakable. This was not the calm assurance of a leader convinced of cruising to victory. It was more the posture of a man who had bet the house whilst the neighbours shout that it’s now on fire.

The scale of the risk and the calm of the capital

OpenAI has committed $1.4 trillion in long-term contracts for energy and computing capacity over the next eight years. Yet it only expects $20 billion in revenue this year. According to HSBC analysts, even if OpenAI's revenues reach $200 billion by 2030, the company will still need an additional $207 billion in funding to remain operational.

And yet…

The people with the money are not retreating. They are doubling down with what may feel like – to anyone with an even slightly nervous disposition - terrifying intensity.

SoftBank has committed a further $22.5 billion in capital to Mr Altman’s cause. Bringing its total expected investment to nearly $35 billion. And Thrive Capital has also continued to buy.

The public markets, meanwhile, could hardly be in sharper opposition. As investors continue to weigh up what’s real and what’s illusory, SoftBank's shares dropped 40% after it sold its entire Nvidia stake. To help fund the OpenAI cause.

The skinny? Softbank is effectively financing OpenAI against the market's collective judgement. So why is this major backer behaving as if the strategy of ChatGPT’s maker is a one-way bet? Why, in fact, are Altman and Masayoshi Son - SoftBank’s founder, chairman and CEO - not panicking? What do they know that we do not?

Suspend disbelief and dig on down

Could it be OpenAI’s work with Jonny Ive, former Apple design supremo? On the face of it, this seems unlikely. But suspend disbelief and let’s dig a little deeper.

Over the past year, the vague mumurs have solidified. And now, we argue, it’s possible to logically deduce what they’re working on. These are the hard constraints they have attested to - an object which is:

• Screenless

• Not designed to be worn

• Intended as a third, unobtrusive AI companion that can be fully aware of its surroundings

  
  

Its aesthetic and emotional goals are to be simple, playful, calming and non-intimidating. Ive jokes about a 'lick test' - an object so inviting that one feels an instinctive urge to touch it.

Altman, meanwhile, contrasts it with the modern smartphone, likening it to walking through Times Square. He says the new device will be a user experience as calm as 'sitting in the most beautiful cabin by a lake and in the mountains'. It is meant for domestic and work spaces. Or travel in a pocket. It is intended to feel personal and owned. Not demanding of attention. The antidote to what Ive’s iPhone has become.

Given the evidence, it is difficult not to conclude that this object will attempt to kickstart an entirely new category: a disembodied, ambient AI – made physical by being encased in gorgeous simplicity.

If this is right, the device will host a presence that maintains state over time. And, as such, the device itself will be a small anchor for something that very much exists independently of it. The product will be less the object. And more what sits within.

The data and processing may live in the cloud, on the device, or drift between both. What will matter is the continuity of the presence you interact with.

Using sensors, wi-fi and Bluetooth this presence would be able to ingest a wide range of information. Or, as we have been casting it in the AI age, personal context. Perhaps it will have access to everything (or at least most things) in our digital universes once we have permissioned it. And it may well comprehend our lives in the physical realm.

So, as well as the data in emails, calendar entries, chats and documents, it could also build a picture of our daily rhythms, work patterns and places. And, via an OpenAI model, it would provide an ever-present presence to talk to. A presence that already understands our situation - and who can take digital action when asked.

Once you take that idea seriously, the competitive landscape might hold the potential to be transformed. And tilted a long way towards OpenAI. If it works – and ChatGPT retains its users ahead of launch - the revenue from such a device may, over time, reach beyond iPhone proportions.

In our current competitive frame, the OS, apps and storage are treated as the main units of context. In a presence frame, the crucial asset may become the disembodied AI that sits between human beings and all other systems.

In that world, Google's and Microsoft's assets still matter. But leverage shifts. The presence layer could be owned by someone else, And that someone else may get to become the primary framing for everything else.

The competitive moats would suddenly look much deeper for OpenAI than today’s.

Seen this way, OpenAI's recent decisions - and Altman's strain - might look different. Are they building the scaffolding for an ambient layer? Is the company investing in a consistent presence that sits above all else?

Even if that is the case, there's no guarantee that this is a strategy that would, or even could, work. But, if this is their genuinely held belief, it would explain the divergence between the public scoreboard and the private valuation.

If you believe you're building the layer through which other software will eventually have to pass, you don’t worry about losing a benchmark. Or even an exclusive partnership with Microsoft. You worry about building the log cabin before anyone else does.

And so, our Context is King series is taking a turn. Before returning to the theme more squarely we shall be exploring this intriguing possibility. In full. Because the more we have considered what a disembodied AI would mean for humanity, the more certain we have become that someone, if not Ive and OpenAI, will build it. And that it will achieve a new level of significance.

Whilst Siri and Alexa have attempted rudimentary forms of disembodied AI before – and arguably Gemini is heading in the same direction – it is difficult to find a serious essay which thinks it all through logically.

So that is what we will do. And we hope you’ll join us along the way.