[DigitalToday reporter Chi-gyu Hwang (황치규)] Big cloud companies are stepping up attacks on the AI chip market dominated by Nvidia. They are now rapidly pushing beyond AI inference and into AI model training, an area where users are widely seen as having little choice but to use Nvidia GPUs despite high costs. AI companies such as OpenAI, Anthropic and Meta appear to be gaining momentum in efforts to use Google and Amazon chips instead of Nvidia GPUs to train their flagship AI models.

Meta recently signed a multiyear contract to rent and use Google’s AI semiconductor, TPU. The deal is reported to be worth several billion dollars. Meta plans to use TPU to train new AI models under the contract. Meta is also in talks with Google on a plan to introduce TPU directly into its data centres from next year.

· Meta signs multibillion-dollar TPU agreement with Google, to use it for AI model training

Amazon is also becoming increasingly aggressive in its push to expand its own AI chip ecosystem. Amazon is investing up to $50 billion in OpenAI and strengthening strategic cooperation, with AI chips included as a core element. Under the cooperation, OpenAI plans to use Amazon Web Services infrastructure and adopt the Trainium AI chip to build a new enterprise platform called Frontier. Chinese AI startup DeepSeek is set to unveil its latest large language model in early March, and is reported to have developed the model using Huawei and Cambricon AI chips.

· Amazon bets $50 billion on OpenAI, boosting push to expand its own AI chip ecosystem · OpenAI raises $110 billion, expands alliance with AWS · DeepSeek’s next-generation multimodal LLM V4 set to debut, developed with Chinese AI chips

These moves across the industry also appear to be affecting Nvidia’s share price. Against that backdrop, Nvidia also appears to be actively pushing back against challengers. Nvidia will unveil a new chip in March specialised for AI inference computing that helps AI models answer user queries. Nvidia has stressed that its GPUs are leaders in both training and inference. The launch of an inference-only chip is seen as a move that takes into account rising demand for cost-efficient inference chips.

· Nvidia growth misses expectations; share price pressured as AI chip competition intensifies · Nvidia to unveil new chip for AI inference computing in March, pushing back against challengers

AI is also the keyword at the Mobile World Congress in Barcelona, Spain. Despite the event being led by telecom companies, strategy and product announcements centred on AI are continuing.

· KT to showcase AI innovation technology combined with K-culture in Barcelona · KT unveils K RaaS robot platform, implementing on-site physical AI · SKT launches full-stack AI, spanning infrastructure, models and services · LG Uplus to showcase human-centred AI technology at MWC26 · Samsung Electronics steps up Galaxy AI push, presenting a future blueprint

Ahead of the opening of MWC, Nvidia unveiled a vision to build 6G based on an AI-native platform together with the global telecom industry. Nvidia aims to transform telecom networks worldwide into AI infrastructure through AI-RAN (artificial intelligence radio access networking) technology.

· Nvidia unveils 6G vision based on AI-RAN technology: "Transform telecom networks into AI infrastructure"

Here is a roundup of moves by companies in and outside South Korea around AI.

OpenAI’s ChatGPT has surpassed 900 million weekly users and is nearing 1 billion. Anthropic’s chatbot Claude is reported to have more than doubled its number of paid subscribers compared with October last year.

· ChatGPT weekly users surpass 900 million, nearing 1 billion · Anthropic’s Claude triples daily sign-ups since November

Google unveiled an AI image-generation model called Nano Banana 2 (Gemini 3.1 Flash Image). Nano Banana 2 can generate images in various aspect ratios from 512px up to 4K resolution, and can depict up to 5 characters and 14 objects. It can also handle complex requests, producing more vivid lighting, textures and detail.

· Google releases Nano Banana 2 model: "Faster and clearer"

As criticism grows that companies are not seeing the expected effects from adopting AI, AI companies are pursuing new strategies. One approach is to spread technology through cooperation with consulting firms. French AI developer Mistral AI also signed a multiyear contract with global consulting firm Accenture.

· Partnerships between leading AI developers and consulting firms spread; Mistral AI also allies with Accenture

Anthropic has moved to restrict the U.S. military’s use of its AI, and its relationship with the U.S. government appears to be growing increasingly uncomfortable. President Donald Trump rejected this and instructed federal agencies to stop using Anthropic products, and Defence Secretary Pete Hegseth designated Anthropic as a supply-chain threat. Even so, Anthropic’s AI chatbot Claude has jumped to No. 1 in Apple’s App Store popularity ranking. In contrast, OpenAI Chief Executive Sam Altman said it signed a contract with the U.S. Department of War (DoW) to deploy its AI models on military networks.

· Trump orders federal agencies to stop using Anthropic AI · Anthropic’s Claude tops App Store rankings amid conflict with Defence Department · OpenAI signs AI supply deal with Defence Department, conditional on "technical safeguards" · OpenAI and Anthropic make divergent choices on national security, drawing criticism of "self-contradiction"

Microsoft is reported to be reviewing the launch of a Microsoft 365 E7 bundle with significantly enhanced AI features. AI platform ServiceNow launched an Autonomous Workforce, an AI specialist that secures the roles, permissions and governance needed for corporate work and performs tasks autonomously. Design software firm Figma will integrate OpenAI’s AI coding tool Codex to strengthen collaboration between designers and developers.

· Microsoft reviews launch of AI-equipped Microsoft 365 E7 bundle · ServiceNow launches AI specialist Autonomous Workforce · Figma expands Codex support in cooperation with OpenAI

Media group Conde Nast, which publishes Vogue and The New Yorker, is preparing a strategy that does not make Google search a core traffic route. Conde Nast CEO Roger Lynch (Roger Lynch) said, "Within a few years, Google search will not play a meaningful role in our traffic."

· Conde Nast CEO: "Google search will not play a meaningful role in traffic within a few years"

AI music generation startup Suno has reached 2 million paying subscribers and annual recurring revenue of $300 million.

· Generative AI music startup Suno reaches $300 million in ARR; paying subscribers surpass 2 million

Samsung Electronics on March 1 announced it will convert its production plants in and outside South Korea into AI-driven factories by 2030. An autonomous factory is one that actively applies AI across the entire manufacturing process, and Samsung Electronics plans to introduce digital twin-based simulations across all processes from materials receipt to production and shipping. It also plans to strengthen data-driven analysis and pre-verification through quality, production and logistics AI agents, and to innovate quality and productivity across its global production hubs. Global open-source solutions company Red Hat launched Red Hat AI Enterprise, an integrated AI platform that can deploy and manage AI models, agents and applications across hybrid clouds.

· Samsung Electronics pushes conversion to AI autonomous factories by 2030 · Red Hat launches integrated AI platform Red Hat AI Enterprise

Physical AI company RLWRLD raised Seed 2 funding to meet rising demand for cooperation with strategic partners in and outside South Korea.

· Physical AI firm RLWRLD raises $26 million in Seed 2 funding, secures many strategic investors

AI payments blockchain company Kite AI officially joined Google’s AP2, an agent payment protocol, as a community partner. WERT Intelligence launched Keywert Insight, an AI technology and research platform based on patent and science and technology data.

· Kite AI joins Google AP2 as community partner · WERT Intelligence launches AI technology and research platform Keywert Insight