• AI KATANA
  • Posts
  • Nvidia revenues jump almost 80% on booming AI chip sales

Nvidia revenues jump almost 80% on booming AI chip sales

Also: Google Unveils AI Scientist That Could Transform Research

Author

AI KATANA
February 27, 2025

Hello!

Today’s edition brings major AI breakthroughs, corporate milestones, and investment shifts shaping the industry. Nvidia’s revenue soared 78% year over year, fueled by surging AI chip demand, but the company faces challenges from competition, US-China trade policies, and next-gen compute demands. Meanwhile, South Korea is building the world’s largest AI data center with a $35 billion investment, positioning itself as a global AI hub. Amazon introduced Alexa+, a new AI-powered assistant with advanced personalization, and Google unveiled its AI Co-Scientist, designed to aid researchers with hypothesis generation and scientific analysis. Microsoft’s Phi-4 models bring powerful AI capabilities to compact, energy-efficient systems, offering an alternative to large-scale models. In corporate insights, Nvidia CEO Jensen Huang predicts AI will require 100x more compute power in the near future, reinforcing the company’s push for more advanced chips. Meanwhile, Alibaba has open-sourced its video-generation AI, providing a free alternative to OpenAI’s Sora, and Disrupt.com pledged $100 million for AI-first startups, signaling continued investment interest despite a downturn in MENA VC funding.

Sliced just for you:

  • 🚀 Nvidia revenues jump 80% on booming AI chip sales

  • 🇰🇷 South Korea secures $35B deal for the world’s largest AI data center

  • 🏠 Amazon unveils Alexa+ and new smart home devices

  • 🧪 Google introduces AI Co-Scientist to assist researchers

  • ⚡ Microsoft’s new Phi-4 models bring big AI power in small packages

  • 🎙️ Nvidia CEO: Next-gen AI needs 100x more compute power

Nvidia’s revenue surged 78% year over year to $39.3 billion, surpassing estimates, while net income jumped 80% to $22.1 billion, driven by soaring demand for its advanced AI chips. The company projects $43 billion in revenue for the next quarter, as data center revenues nearly doubled due to Big Tech’s rapid AI expansion. Despite concerns over competition from China’s DeepSeek, which claims to train AI models on less advanced chips, CEO Jensen Huang emphasized that demand remains strong for Nvidia’s latest-generation Blackwell chips. Some early production issues, including overheating reports, were noted, but the transition is progressing smoothly. However, Nvidia faces potential challenges from evolving US export controls and tariffs, which could impact its position in the escalating AI rivalry between Washington and Beijing. While its stock has rebounded from a sharp drop last month, Nvidia is no longer the explosive force it once was in driving US markets, as uncertainties around regulations and geopolitical tensions persist.

South Korea’s Jeollanam-do Province has signed a $35 billion deal with Fir Hills (FH), a subsidiary of Stock Farm Road (SFR), to construct the world’s largest AI data center. This 3GW facility will generate 10,000 new jobs and position Jeollanam-do as a global AI hub. The project is backed by a strategic partnership with Stanford University, focusing on AI, energy efficiency, and regulatory innovation, led by Professor Michael Lepech. Key advisors include John Hennessy, chairman of Alphabet Inc., and Neelie Kroes, former VP of the European Commission, ensuring compliance with international digital standards. The initiative is also supported by KENTECH University, a South Korean energy research institution aiming to lead in sustainable energy solutions. Construction is set to begin in winter 2025, solidifying South Korea’s growing presence in AI infrastructure and digital innovation.

Amazon’s 2025 Devices & Services event introduced Alexa+, a next-generation AI-powered assistant designed to enhance user experiences through more natural conversations, deep personalization, and advanced automation. Alexa+ integrates with thousands of services, leveraging Amazon’s Bedrock platform and models from Anthropic to provide dynamic assistance, from booking services to generating creative content. It costs $19.99 per month but is free for Prime members. New Echo Show 21 and Echo Show 15 devices were also unveiled, featuring adaptive displays and smart home controls. Alexa+ now enables more intuitive home automation, seamless multi-room music control, and direct document-based queries. New kid-friendly features, “Stories with Alexa” and “Explore with Alexa,” were introduced to foster creativity and learning. Additionally, Ring unveiled new security cameras with Smart Video Search and 360-degree views. Amazon CEO Andy Jassy emphasized the company’s long-term AI vision, stating that generative AI will redefine customer experiences across all industries.

Google has introduced the AI Co-Scientist, a groundbreaking AI tool built on its Gemini 2.0 model, designed to assist researchers in forming hypotheses, analyzing scientific literature, and generating structured research proposals. Unlike conventional AI chatbots, this system mimics the scientific method by using multiple specialized AI agents—such as Generation, Reflection, and Evolution—to iteratively refine ideas. The AI Co-Scientist has already demonstrated success in biomedical research, identifying potential drug repurposing candidates for acute myeloid leukemia, discovering anti-fibrotic targets for liver disease, and providing insights into antimicrobial resistance. Google is launching the tool under a Trusted Tester Program, with contributions from DeepMind, Google Cloud AI, and leading academic institutions. While the AI offers significant potential, Google acknowledges challenges in factual accuracy, external validation, and data security concerns, particularly regarding research data privacy. Rather than replacing human scientists, the AI Co-Scientist is designed to augment research efficiency, allowing scientists to focus on critical thinking and innovation while the AI handles data-heavy tasks.

Microsoft has unveiled the Phi-4 AI models, a new class of small yet powerful AI systems designed for efficiency, multimodal processing, and deployment on standard hardware. The Phi-4-Multimodal model (5.6 billion parameters) and the Phi-4-Mini model (3.8 billion parameters) outperform similarly sized models and even rival systems twice their size in text, image, and speech processing. The “Mixture of LoRAs” technique allows Phi-4-Multimodal to integrate multiple input types seamlessly while maintaining accuracy. These models are optimized for low-power computing environments, enabling AI applications to run efficiently on edge devices instead of requiring cloud-based processing. Phi-4-Mini excels in math, coding, and reasoning, scoring 88.6% on the GSM-8K benchmark and 64% on the MATH benchmark, surpassing larger models. Microsoft aims to democratize AI access by making Phi-4 available through Azure AI Foundry, Hugging Face, and the Nvidia API Catalog, positioning it as an alternative to large-scale, resource-intensive AI models.

🎙️ Nvidia CEO Huang says AI has to do ’100 times more’ computation now than when ChatGPT was released

Nvidia CEO Jensen Huang emphasized that next-generation AI will require 100 times more compute power than previous models due to advanced reasoning techniques that break down problems step by step. Speaking after Nvidia’s 78% revenue surge to $39.3 billion, he highlighted models like DeepSeek’s R1, OpenAI’s GPT-4, and xAI’s Grok 3as key examples of this shift. Huang dismissed concerns that Chinese firms could reduce AI infrastructure costs, arguing that reasoning models will demand more powerful chips, not fewer. Nvidia’s data center revenue skyrocketed 93%to $35.6 billion, reinforcing its dominance in AI hardware. However, US export restrictions have halved Nvidia’s China revenue, while Huawei’s growing presence adds competitive pressure. Despite these challenges, Huang stated that software will find a way to optimize AI performance, regardless of hardware limitations. Nvidia’s GB200 chip, designed for the US market, can generate AI content 60 times faster than export-restricted versions sold in China.

🛠️ AI tools updates

Alibaba has open-sourced its video-generation AI model, Wan2.1, offering four versions designed for high-quality text-to-video generation. These models range from 1.3 billion to 14 billion parameters, producing videos up to 720p resolution, making them a free alternative to OpenAI’s Sora and Google’s Veo 2. Wan2.1 is optimized for realistic visuals, complex motion handling, and enhanced pixel quality, adhering to physical principles for more accurate video generation. Analysts highlight the growing enterprise interest in video AI, though concerns over deepfakes and security risks remain. The models are available on Alibaba Cloud’s Model Scope and Hugging Face, potentially expanding to other cloud platforms, similar to DeepSeek’s AI chatbot. While still in its early stages, AI-driven video generation is expected to become a major market, especially for creative, media, and business applications.

💵 Venture Capital updates

UAE-based venture builder Disrupt.com has pledged a $100 million investment to support AI-first startups across AI, cybersecurity, Web 3.0, automotive technology, and retail innovation. Founded by Aaqib, Uzair, and Umair Gadit, who previously exited Cloudways for $350 million, the firm has already deployed $40 million across growth-stage and early-stage companies. Unlike traditional venture capital firms, Disrupt.com follows a “CoBuild” model, acting as a fractional co-founder by providing engineering, go-to-market, and operational support to startups. The company focuses on pre-seed to Series A investments, prioritizing profitability and product-market fit over unchecked growth. Disrupt.com has already backed AI-driven startups like Agentnoon (an organizational transformation platform) and Ahya (a climate action tool). The announcement comes amid a 29% drop in MENA VC funding, positioning Disrupt.com as a key player in revitalizing AI innovation in the region.

🫡 Meme of the day

⭐️ Generative AI image of the day