Who Wins When Robots Fight Robots?
Sign up for ARPU: Stay ahead of the curve on tech business trends.
Cybersecurity's Kafka Problem
Modern cybersecurity has entered a truly Kafka-esque state of affairs. The problem, it seems, is artificial intelligence. The solution, it turns out, is also artificial intelligence. Last week, identity management platform Okta and cybersecurity company SentinelOne provided a financial snapshot of just how profitable this AI-on-AI cycle has become. Both companies raised their revenue forecasts on the back of surging demand for their AI-powered security tools. The gold rush is not just because their defensive AI is good; it's because the attackers' offensive AI is also getting very good, very fast. This has created a strange and lucrative new theater of war, one fought almost entirely between autonomous systems.
As SentinelOne CEO Tomer Weingarten articulated in their Q4 2025 earnings call:
...AI is no longer experimental. And in the hands of attackers, it's a real threat. The scale, automation, and speed of attacks are accelerating...Our industry cannot afford to rely on the same outdated approaches of the past two decades. They simply don't work. Our AI native autonomous security is fundamentally redefining how cybersecurity challenges are addressed, setting us apart in the industry.
…we're the first company to embed foundational generative AI capabilities into every platform solution by default, end point, cloud, data solutions, and more. From the beginning, we introduced an AI-based approach to endpoint security.
In a nutshell, the nature of cybersecurity threat has fundamentally changed. The battle is no longer against a human hacker who has to manually find vulnerabilities; it’s against an automated, AI-powered adversary that can launch attacks at machine speed, 24/7. The same generative AI that can write marketing copy can also write highly convincing phishing emails at an unprecedented scale. The only way to fight an AI attacker is with an AI defender.
This is the core pitch of companies like SentinelOne. Their AI systems ingest billions of data points across a network in real-time, learn what "normal" behavior looks like, and can automatically detect and neutralize new, never-before-seen threats in seconds. This is a task far beyond the capacity of any human security team.
The whole thing is creating an absurd, self-feeding growth cycle. The more businesses integrate AI into their own operations, the larger their digital footprint—or "attack surface"—becomes. This, in turn, creates more vulnerabilities for AI-powered attackers to exploit, which then drives even greater demand for AI-powered defense. It is a perpetual, escalating loop where the solution to the problem created by AI is, inevitably, more AI.
The Enterprise Bottleneck
The thing about the AI revolution is that while chatbots have taken over the consumer world, the corporate world has been surprisingly resistant. For all the talk of transformation, the brutally difficult work of integrating AI into the core software that runs a business—finance, HR, supply chain—has been painfully slow. This is the enterprise bottleneck. It is the biggest, most boring, and most lucrative problem in AI right now. And a strange new alliance between Oracle and Google shows how the arms race is shifting to solve it.
Earlier this month, Oracle announced it would begin offering AI models from its archrival, Google, on its own cloud platform. This is a little strange. It’s like Ford agreeing to sell GM engines in its F-150s. But the deal is not really about cloud infrastructure. It’s about a strategic realignment to overcome the enterprise bottleneck, a challenge that even tech giants are finding difficult to solve alone.
One way to think about it is that the AI model makers and the enterprise software incumbents each hold different, critical pieces of the puzzle.
Google has one of the world's most powerful AI "brains" in Gemini, but it doesn't own the "hands" or the "central nervous system" inside most large corporations. It can offer a brilliant model, but it has no easy way to plug it into the core financial workflows of a Fortune 500 company.
Oracle, on the other hand, is the central nervous system. Its Fusion Cloud and NetSuite applications are the backbone for thousands of the world's biggest businesses. What it lacks is a homegrown, frontier AI model that can compete with the best from Google or OpenAI.
This partnership is a powerful arrangement between two formidable enterprise players. Oracle gets to arm its massive customer base with a best-in-class AI model, giving them a compelling reason to stay within its ecosystem. Google gets a strategic channel to embed its premier AI model into the high-value corporate workflows that have, until now, been largely out of its reach.
This Oracle-Google partnership doesn't exist in a vacuum; it’s a direct response to a broader industry race where different philosophies are emerging to solve the same enterprise problem:
- The Microsoft Bloc: The clear front-runner, Microsoft’s true advantage lies less in its deep investment in OpenAI and more in its near-monopoly on the digital real estate where work actually happens. By controlling the enterprise operating system (Windows) and the indispensable productivity suite (Office), it can embed its Copilot assistant seamlessly into the daily workflow of billions. This makes Copilot the default AI for work, a powerful incumbent advantage in the race for enterprise adoption.
- The Amazon Bloc: As the undisputed leader in cloud infrastructure, Amazon is pursuing a classic platform play. Its strategy is less about pushing a single integrated AI assistant and more about making AWS the indispensable foundation where enterprises can build their own AI solutions. It offers the powerful models from its strategic partner, Anthropic, as a premier option, but is also happy to rent its vast cloud power to customers using models from other providers. The bet is to win by owning the underlying infrastructure, regardless of which AI 'brain' the customer chooses.
- The Salesforce Approach: Salesforce’s recent acquisition of data-management firm Informatica highlights a different strategy. The bet here is that a model is only as good as the data it can access, and that winning the enterprise means solving the complex data integration problem first.
The battle for AI supremacy is moving away from the consumer chatbot and into the unglamorous but incredibly valuable world of enterprise resource planning. It won't be won by the model with the best benchmark scores, but by the ecosystem that can most successfully embed its intelligence into the boring, essential plumbing of global business.
The Scoreboard
- Semiconductor: Nvidia’s top two mystery customers made up 39% of the chipmaker’s Q2 revenue (CNBC)
- AI: China Has a Different Vision for AI. It Might Be Smarter (WSJ)
- Infra: Dell slides as high AI server costs, competition blunt upbeat demand forecast (Reuters)
Enjoying these insights? If this was forwarded to you, subscribe to ARPU and never miss out on the forces driving tech: