AI upheavals: platforms, pricing, and new players redraw the competitive map.

Controlling distribution has become as critical as the AI technology itself. Meta’s launch of its AI assistant *Muse* underscores the power of platform scale: in just 12 days, the mobile app amassed 2.8 million downloads – more than OpenAI’s ChatGPT managed in the same post-launch period ([1]). By leveraging its social media and messaging empire, Meta propelled Muse to the top of app store charts almost overnight.

Notably, Meta’s success with Muse was not due to any radical new AI capability, but rather aggressive cross-promotion and ecosystem integration. The company repeated the playbook used with its Threads app (which hit 500 million users via tie-ins with Instagram and Facebook) ([2]), embedding Muse across its platforms. Over 95% of early Muse users were already Facebook account holders, and 63% were Instagram users ([3]) – a vivid reminder that an existing user base can fast-track a new AI service’s adoption. In effect, distribution channels and data access are acting as key moats: an incumbent like Meta can instantly give its AI offering reach that a standalone player cannot match.

But platform power cuts both ways, as shown by Amazon’s defensive move against that very same AI agent. This week Amazon began outright blocking Muse from making any purchases on Amazon.com, effectively slamming the door on Meta’s foray into its shopping platform ([4]). By Sunday night, any user trying to buy products via Muse was greeted with a warning that the “AI agent” violated Amazon’s terms of service – essentially labeling Muse as an unwelcome infiltrator on the site. It’s the first time a major online retailer has explicitly barred a mainstream AI assistant as an “unauthorized” user ([5]), setting a new precedent in how platform owners can exert control.

Amazon’s blockade highlights that whoever owns the platform can dictate the rules in an AI-driven economy. The e-commerce giant’s logic is both defensive and strategic. A third-party shopping agent threatens to come between Amazon and its customers, potentially bypassing Amazon’s storefront and even undercutting its $68 billion online advertising business that relies on users browsing the site ([6]). Amazon also raised immediate concerns about Muse’s transparency and security – noting that Meta hadn’t formally permissioned the bot, which was not clearly identified as non-human and might capture user credentials without consent ([7]). If the AI helper makes incorrect orders, Amazon would still be on the hook to resolve customer service issues. In short, Amazon has every incentive to guard its platform until it can introduce its own AI shopping assistant on its own terms. The broader lesson for leaders in any sector is clear: controlling a large platform or distribution channel can be as important as algorithmic innovation, since it allows incumbents to accelerate their own AI offerings – or shut out those of rivals.

Incumbents are also betting that dedicated devices can extend their AI advantage. On September 21, Google opened pre-orders for its first AI-centric laptop – the $899 “Googlebook” – which runs on a custom Android OS and deeply embeds Google’s upcoming *Gemini* AI model into the user experience ([1]). This is not a standard Chromebook upgrade; the Googlebook introduces system-level AI features like a "magic" cursor that can interpret on-screen content, AI-driven voice dictation cleanup (codenamed Rambler), and context-aware widgets – all powered by Gemini’s latest capabilities ([2]). The device even bundles a year of Google’s AI Pro subscription (including extra cloud storage and advanced AI tools) to lock users into the ecosystem from day one ([3]).

Google’s move is a bold attempt to change the basis of competition in personal computing. Rather than relying on pure software updates, it is asking consumers – and especially schools – to invest in new hardware for a more seamless AI experience. The company explicitly frames the Googlebook as a step up from the 50 million-plus Chromebooks already used in schools ([4]), signaling a long-term play to migrate a generation of students and educators onto an AI-native platform. If successful, this strategy could strengthen Google’s grip on both the education market and the broader computing ecosystem by making its AI a built-in, indispensable feature of daily workflows.

The Googlebook also underscores a potential shift in consumer expectations: AI capabilities may become a deciding factor in hardware purchases. Competitors like Microsoft (which has reportedly partnered with AI firm *HUMAIN* on an "AI PC" for enterprises) and Apple (weaving more intelligence into devices and assistants) are unlikely to sit still. Senior executives should watch this space – if AI-driven hardware gains traction, companies will need to consider how integrating bespoke AI into products (or partnering to do so) can be a source of sustainable differentiation, and what it means for existing hardware and software roadmaps.

While tech giants play to their strengths, a new generation of AI-native startups is directly targeting incumbents in specific industries. One recent example is *Tabby*, an AI-driven bookkeeping platform created by a former accountant to automate the work of human bookkeepers and replace small-business accounting software ([1]). Tabby links to companies’ financial accounts and uses AI to generate real-time dashboards of their financial health, effectively handling tasks like transaction categorization and tax preparation behind the scenes ([2]). As founder Ahad Ali observed, many small businesses "don’t need better accounting software… They need something that just does it for them" ([3]) – a pain point his seven-person startup aims to solve by making bookkeeping an invisible, AI-provided service rather than a tedious process for customers.

In just 14 months of beta, Tabby has signed up 5,500 small-business users and reached about $100,000 in annual recurring revenue with a lean team ([4]). It’s positioning itself as an alternative to Intuit’s QuickBooks – the long-dominant accounting software for SMEs – by charging a simple monthly fee to handle all bookkeeping tasks autonomously. The competition is heating up: established players like QuickBooks now face not only each other but also well-funded newcomers such as Sequoia-backed *Rillet* and even AI offerings from major tech labs aiming to manage finances ([5]). The common thread is that these disruptive startups promise to remove complexity (and labor costs) from business processes, effectively moving traditional B2B software into an "AI-as-a-service" model.

For C-level strategists, the rise of niche AI disruptors like Tabby is a warning shot. Tasks and services once considered safe behind a software license or skilled staff can suddenly be taken over by a small AI startup with a fraction of the resources. Executives should identify which parts of their value chain are vulnerable to end-to-end AI automation. The winners in the next phase will be those who either preemptively reinvent their own offerings with AI or acquire/partner with the disruptors – before customers opt for a solution that simply “does it for them.”

Another critical shift is the rapid commoditization of AI capabilities, led by aggressive moves from new players. In the last two days, Shanghai-based lab *StepFun* unveiled *Step 5*, a 600-billion-parameter AI model that approaches frontier-level performance at a fraction of the usual cost ([1]). This massive model employs a sparse "mixture-of-experts" architecture and offers a 1 million token context window, enabling it to tackle very large inputs (like entire codebases or lengthy legal documents) in a single go. Most striking is Step 5’s pricing: about $1 per million input tokens (and $2.70 per million output tokens), roughly **7×** cheaper than OpenAI’s top-tier models with similar capabilities ([2]). Moreover, StepFun plans to openly release the model’s weights on October 15, 2026, effectively open-sourcing a system that rivals the best models from Google, OpenAI, and others ([3]).

Industry analysts see this as a major escalation in the AI arms race, shifting competition toward price and openness. Step 5 reportedly delivers around 90% of the performance of leading U.S. models at only ~14% of the cost ([4]). If those figures hold, it could dramatically lower the cost of implementing advanced AI, putting pressure on established providers (which have enjoyed premium pricing) to respond or risk users flocking to cheaper, open alternatives. Already, other big players are reacting: Alibaba’s cloud division just introduced its own 1-million-context *omnimodal* model with certain AI services priced over 98% lower than before – aiming to undercut global rivals and attract usage on its platform.

The broader implication is that advanced AI is rapidly becoming a commodity, eroding one pillar of competitive advantage for market leaders. Proprietary models that once seemed unassailable are now facing challengers willing to nearly give away equivalent capabilities. For business leaders, this means the basis of competition in AI is shifting from purely who has the most advanced model to who can integrate AI most effectively, control access to critical data and customers, and perhaps offer the best economics. Companies must prepare for AI-driven price pressure and consider how to differentiate when cutting-edge AI becomes ubiquitous and inexpensive.