AI Power Plays: New Models, New Rivals, New Rules

([1])ByteDance (TikTok's parent company) has unveiled a powerful new AI model as it expands beyond social media and into advanced AI platforms. Named Doubao 2.0 Pro, this large-scale AI assistant is designed for the 'agent era', capable of complex, multi-step tasks and autonomous decision-making beyond simple Q&A ([2]). The company claims its performance rivals top Western systems like OpenAI's GPT-5.2 and Google's latest Gemini – all while cutting usage costs by roughly an order of magnitude ([3]).

([4])The Chinese tech giant's push is already backed by enormous scale. Doubao now leads China's AI chatbot market with 155 million weekly active users ([5]) – nearly double the users of its closest domestic competitor (DeepSeek, at 81.6 million). Established rivals are scrambling: in one recent move, Alibaba poured 3 billion yuan (~US$548 million) into promotions for its Qwen AI chatbot, briefly vaulting Qwen's daily users from 7 million to 58 million ([6]). This aggressive competition shows that Chinese firms view ubiquitous AI assistants as strategic platforms – worth massive investment to dominate.

Strategically, ByteDance's foray into advanced AI services signals that incumbent consumer-tech platforms can rapidly evolve into AI powerhouses. By leveraging its huge user base and offering high-end AI capabilities at a fraction of the typical cost ([7]), ByteDance is already eroding the early lead of U.S. AI providers in key domains. Business leaders should note that new AI titans – armed with scale and cost advantages – can emerge from unexpected places, forcing a recalibration of global partnership and innovation strategies.

([1])In the United States, four-year-old startup Baseten has raised an eye-popping $1.5 billion at a $13 billion valuation. This massive round – Baseten's fourth in 18 months – highlights the intense investor appetite for companies that supply the infrastructure underpinning the generative AI boom.

([2])Baseten provides software and cloud infrastructure for companies to deploy their own AI models, positioning itself as a cheaper alternative to relying on Big Tech providers like OpenAI or Anthropic ([3]). The company says its revenue has surged 20-fold over the past year as demand soars for “inference” services that power real-world AI applications ([4]).

One investor described Baseten's rise as 'the biggest shift that we've seen in both unit economics and competitive leverage within the AI market so far' ([5]). By enabling organizations to run cutting-edge models on affordable infrastructure, firms like Baseten – and the open-source AI movement it supports – can loosen the grip of incumbent providers. Enterprise clients gain new options beyond the major cloud platforms, compelling established players to rethink their pricing and integration strategies.

([1])Major AI incumbents are also moving quickly to solidify their leads. This week OpenAI launched GPT-5.5-Cyber, a specialized version of its flagship model tailored for cybersecurity defense and software vulnerability analysis ([2]). In testing, this model scored 85.6% on a critical security benchmark – outperforming the standard GPT-5.5's 81.8% on the same test ([3]).

Unlike prior releases, OpenAI is not making its latest AI broadly available. Access to GPT-5.5-Cyber is being limited to a select cohort of vetted partners and government-affiliated defense organizations ([4]). Through the new Daybreak program, OpenAI has aligned with leading cybersecurity firms (including Accenture, CrowdStrike, and IBM) to integrate this powerful model into their products and national cyber defense systems ([5]). By combining exclusive access with strategic alliances, OpenAI is both addressing safety concerns and cementing its influence in critical infrastructure sectors.

At the same time, the biggest tech companies are pouring unprecedented resources into AI infrastructure. The combined AI-related capital expenditures of Amazon, Alphabet, Microsoft, and Meta could top $770 billion in 2026 ([6]) – turning scale itself into a competitive advantage. Even non-traditional players are entering the fray: Elon Musk's SpaceX has built “Colossus” AI supercomputers and just signed a multi-year deal worth up to $6.3 billion to supply advanced Nvidia chips to open-source startup Reflection AI ([7]). And oil giant Chevron announced a 20-year plan to construct a 2.67-gigawatt power plant dedicated to Microsoft's AI data centers in Texas ([8]). These massive investments in compute and energy illustrate how incumbents are fortifying the foundations of AI – raising barriers to entry and creating new “moats” that late-moving competitors will struggle to cross.

([1])Meanwhile, a new wave of AI solutions is challenging the idea that one model can do it all. Japan's Sakana AI has introduced Fugu, a “multi-agent” system that coordinates numerous specialized AIs behind a single OpenAI-compatible API ([2]). Instead of depending on one large model, Fugu dynamically orchestrates a team of expert models for different tasks – letting clients tap multiple best-in-class AI skills via one interface, without locking into any one vendor ([3]).

([4])This multi-model approach is already delivering results at the frontier of performance. Fugu reportedly matches or exceeds the performance of leading proprietary models like OpenAI’s GPT-5.5, Google’s Gemini 3.1, and Anthropic’s Claude Opus 4.8 on key coding, reasoning, and STEM benchmarks ([5]). By assigning the optimal model to each sub-task, an orchestrated AI system can achieve top-tier outcomes without any single company having a monopoly on the best model.

([6])At the same time, open-source AI models continue to rapidly narrow the gap with Big Tech's offerings. The newly released GLM-5.2 (a 744B-parameter open model from Zhipu AI) now ranks as the highest-performing publicly available model on a leading industry index, placing just behind the top two proprietary systems ([7]). As freely available models approach “good enough” capabilities, they empower startups and enterprises to innovate on their own terms – pressuring established players to provide more than just superior algorithms to maintain a competitive edge.

([1])Another emerging theme is how incumbents that once fought AI are now finding value in partnership. Getty Images, a global stock-photo leader, spent years resisting generative AI – banning AI-generated art and even suing an AI startup over alleged copyright infringement (a case that was dismissed) ([2]). Now Getty has pivoted to collaborate: it recently launched its own AI image-generator trained on Getty's licensed library, and this week announced a landmark deal to integrate its content into OpenAI's ChatGPT ([3]).

([4])The market reacted swiftly to Getty's change of heart. Its stock price surged as much as 145% on the day the OpenAI partnership was unveiled ([5]). The multi-year agreement will let ChatGPT's vast user base access high-quality, rights-cleared images within AI chats – with Getty's CEO noting that licensed visuals make AI results more trustworthy for users ([6]) ([7]). For incumbents in data-rich industries, the lesson is clear: embracing strategic alliances with AI platforms (or developing proprietary AI services on their own data) can turn a disruptive threat into a mutual growth opportunity.