Microsoft Faces Copilot Sales While Meta Llama Sees Delays

The artificial intelligence sector is experiencing intense competition and rapid advancements, with Microsoft cofounder Bill Gates predicting a "hyper competitive" future where many current high-valuation AI companies may not endure. This environment fuels a 21st-century tech arms race, as nations like the U.S., China, and Russia integrate AI into military capabilities, from robotic aircraft to AI-designed bioweapons. Despite the rapid progress, AI experts at the NeurIPS conference admit a fundamental challenge: no one fully understands how leading AI models truly operate, highlighting issues with "interpretability" and current evaluation methods.Meta is navigating a significant shift in its AI strategy, facing internal confusion and delays. Its new frontier AI model, codenamed Avocado, originally slated for late 2025, is now expected in early 2026 due to training performance issues. This follows the underperformance of Meta's Llama 4 model and a move away from open-source AI, with Avocado potentially becoming proprietary. CEO Mark Zuckerberg has restructured the AI division, replacing Chris Cox with Alexandr Wang as the new chief AI officer. Meta also acquired Limitless and is adjusting its metaverse budget and mixed-reality glasses timeline.Meanwhile, Microsoft faces challenges in selling its AI products, including Copilot, which holds 14.1 percent of the market. Another unnamed AI tool leads with 61.3 percent, and Google's Gemini is closely behind at 13.4 percent, projected to surpass Copilot soon with 12 percent quarterly user growth compared to Copilot's 2 percent. Google, in partnership with MediaTek, is boosting on-device AI capabilities with the LiteRT NeuroPilot Accelerator. This technology allows large language models like Qwen3 and Gemma to run directly on MediaTek Dimensity NPUs in various devices, reducing reliance on data centers.Internationally, Qatar launched Qai on December 8, 2025, a national AI company under its $524 billion sovereign wealth fund. Qai plans to invest in AI infrastructure globally and develop "trusted" AI systems, focusing on commercializing existing models rather than building its own. The U.S. is also strengthening AI partnerships, with President Donald Trump visiting the Arab Gulf to secure deals. The U.S. agreed to sell advanced chips to Saudi Arabia and the UAE, while Gulf countries pledged billions in U.S. AI investments, aiming to make the region a major AI computing hub.However, these international collaborations are subject to new U.S. export controls on high-performance AI chips. The U.S. government requires approval or prohibits the export of sensitive AI technology, even if made by U.S. companies abroad, to prevent diversion to adversaries for military use. The Bureau of Industry and Security (BIS) recently issued guidance for due diligence on exports to regions like Macau and China. The U.S. House is also considering the GAIN AI Act, which would prioritize U.S. customers for these critical chips.In practical applications, Sabre Corporation introduced SabreMosaic Concierge IQ, an AI chat solution for airlines to enhance traveler planning and management. Brook.ai received a Gold Award for its AI-powered remote care management, aiming to integrate continuous care into U.S. healthcare. The increasing presence of "rogue" AI agents is also prompting companies, as noted by SailPoint CEO Mark McClain, to urgently tighten data security to protect sensitive information from misuse.

Key Takeaways

• Bill Gates predicts the AI industry will become "hyper competitive," with many current high-valuation companies likely to fail.
• Meta's new frontier AI model, Avocado, is delayed until early 2026 and may be proprietary, following Llama 4's underperformance and a leadership change with Alexandr Wang as chief AI officer.
• Microsoft's Copilot holds 14.1% of the AI market, but Google's Gemini, with 13.4%, is projected to surpass it due to higher user growth.
• Google and MediaTek introduced the LiteRT NeuroPilot Accelerator, enabling large language models like Qwen3 and Gemma to run directly on MediaTek Dimensity NPUs for on-device AI.
• AI experts admit they do not fully understand how leading AI models work, highlighting challenges in "interpretability" and evaluation.
• Qatar launched Qai on December 8, 2025, a national AI company under its $524 billion sovereign wealth fund, to invest in AI infrastructure and commercialize existing models.
• The U.S. is partnering with Gulf states, selling advanced AI chips to Saudi Arabia and the UAE, while Gulf countries commit billions to U.S. AI investments.
• New U.S. export controls on high-performance AI chips require approval for sensitive technology exports and due diligence to prevent diversion to adversaries.
• Sabre Corporation launched SabreMosaic Concierge IQ, an AI chat solution for airlines to improve traveler planning, booking, and management.
• The rise of "rogue" AI agents is driving companies to urgently tighten data security measures to protect sensitive information.

Qatar establishes new national AI company Qai

Qatar launched a new national artificial intelligence company called Qai on December 8, 2025. Qai will operate under the Qatar Investment Authority, the country's $524 billion sovereign wealth fund. The company plans to invest in AI infrastructure both in Qatar and abroad, offering high-performance computing and a suite of AI tools. Abdulla Al-Misnad, who chairs Qai, stated the firm will focus on developing "trusted" AI systems. Unlike some other Gulf states, Qai will not build its own large language models but will evaluate and commercialize existing ones and work with new technologies like autonomous agents.

US and Gulf partner on AI technology

U.S. President Donald Trump visited the Arab Gulf in May, focusing on artificial intelligence deals. The U.S. agreed to sell advanced chips to Saudi Arabia and the UAE and invest in AI mega-campuses there. In return, Gulf countries pledged to invest billions in U.S. AI. Saudi Crown Prince Mohammed bin Salman recently secured final approval for thousands of U.S. semiconductors. This partnership aims to make Gulf states a major AI computing hub, channel their wealth into American AI companies, and expand U.S. AI influence globally. Daniel Benaim highlights the need for strict safeguards to prevent technology misuse.

New US rules impact AI chip exports

New U.S. export controls on high-performance AI chips are creating challenges for data center operators outside the U.S. The U.S. House of Representatives is considering the GAIN AI Act, which would prioritize U.S. customers for these chips. The U.S. government requires approval or prohibits the export of sensitive AI technology, even if made by U.S. companies abroad. The Bureau of Industry and Security (BIS) is concerned about chips being diverted to U.S. adversaries for military use. BIS recently issued new guidance, requiring due diligence for exporters of advanced computing integrated circuits to places like Macau and China.

Meta changes AI strategy with new Avocado model

Meta's artificial intelligence strategy is shifting, causing internal confusion as the company tries to keep up with rivals like OpenAI and Google. Meta is developing a new frontier AI model, codenamed Avocado, which may be proprietary instead of open source. Originally expected in late 2025, Avocado's launch is now planned for the first quarter of 2026 due to training performance issues. This change follows the underperformance of Meta's Llama 4 model. CEO Mark Zuckerberg has reorganized the AI division, replacing Chris Cox with Alexandr Wang as the new chief AI officer.

Meta changes AI and hardware plans amid delays

Meta Platforms Inc. is restructuring its artificial intelligence and hardware strategies due to delays with its new Avocado AI model and internal leadership changes. CEO Mark Zuckerberg has shifted away from open-source AI, and Avocado, now potentially proprietary, is delayed until early 2026. This led to Chris Cox's removal from the AI division and Alexandr Wang's appointment as chief AI officer. Meta also acquired Limitless, an AI wearable tech startup, and is cutting its 2026 metaverse budget by up to 30 percent, delaying Phoenix mixed-reality glasses to early 2027. The company is also forming cloud partnerships with CoreWeave Inc. and Oracle Corp.

AI experts admit they do not understand how it works

At the NeurIPS conference in San Diego, artificial intelligence experts made a surprising admission: no one fully understands how leading AI models work. Researchers are grappling with "interpretability," the challenge of figuring out why these complex systems behave as they do. Google's team is moving towards more practical goals, while OpenAI's team still aims for ambitious reverse-engineering. Experts also noted that current evaluation methods are not good enough for measuring complex AI behaviors like intelligence. Despite these challenges, researchers remain hopeful about AI's potential to advance scientific research.

Sabre launches AI chat tool for airlines

Sabre Corporation launched SabreMosaic Concierge IQ, a new artificial intelligence chat solution for airlines. This tool helps travelers plan, book, and manage their trips through simple conversations. It aims to improve the traveler experience by offering personalized offers and support throughout their journey. SabreMosaic Concierge IQ integrates with current airline systems, allowing for smooth communication and transactions. This launch shows Sabre's dedication to using AI to innovate in travel technology.

Microsoft faces challenges selling AI products

Microsoft is reportedly having difficulty selling its artificial intelligence products, including Copilot. While Copilot holds 14.1 percent of the market, another unnamed AI tool leads with 61.3 percent, and Gemini follows closely with 13.4 percent. Gemini is expected to surpass Copilot soon, with an estimated 12 percent quarterly user growth compared to Copilot's projected 2 percent. Despite AI agents sometimes creating false information, their overall use is expected to continue growing.

Brook.ai wins top award for AI remote care

Brook.ai received a Gold Award for Remote Patient Monitoring at the 27th annual Digital Health Awards. The Health Information Resource Center organized this prestigious program. Brook.ai is recognized for its pioneering work in AI-powered remote care management. The company aims to make remote care a key part of U.S. healthcare, extending continuous care beyond traditional doctor's offices.

America faces new tech arms race

The world is in a 21st-century arms race driven by rapid advances in artificial intelligence, synthetic biology, and quantum computing. These technologies are transforming warfare, with potential weapons like robotic aircraft swarms and AI-designed bioweapons. The United States leads in some areas, especially AI, thanks to private sector investments. However, countries like China and Russia are quickly investing state resources to integrate these innovations into their militaries. To keep up, America needs strong political will and close teamwork between the government, private companies, and research institutions.

Bill Gates predicts fierce competition in AI

Bill Gates, cofounder of Microsoft, warned that the artificial intelligence industry will become "hyper competitive." He believes many AI companies with high valuations today will not succeed in the long run. Gates emphasized that AI is the "most important thing going on" and a "deeply profound technology" that will reshape the world. He remains confident that AI will bring significant benefits, especially in health, education, and agriculture. Gates predicted that next year will be big for global health, with new AI tools like virtual doctors and farm advisors being tested, particularly in Africa to boost productivity.

Rogue AI agents make companies secure data

The increasing presence of "rogue" artificial intelligence agents is prompting companies to tighten their data security. SailPoint founder and CEO Mark McClain explained that businesses are quickly working to protect sensitive information. This urgent need comes as artificial intelligence technology continues to expand rapidly. Companies aim to prevent unauthorized access or misuse of their data by these autonomous AI systems.

Google and MediaTek boost on-device AI

Google and MediaTek introduced the LiteRT NeuroPilot Accelerator, a new technology that allows large language models to run directly on devices. This innovation targets MediaTek Dimensity NPUs in phones, laptops, and IoT hardware, reducing the need to send every AI request to a data center. LiteRT, the successor to TensorFlow Lite, now directly connects with MediaTek's NeuroPilot NPU stack, offering developers a single way to deploy AI models. It supports various open-weight models like Qwen3 and Gemma, making on-device AI more efficient and accessible.

Sources

• Qatar Launches National AI Firm as Gulf Tech Investments Ramp Up
• Compute Is the New Oil
• AI Chip Export Controls: A New Challenge for Data Centers
• From Llamas to Avocados: Meta's shifting AI strategy is causing internal confusion
• Meta Resets AI, Hardware Plans As 'Avocado' Delays, Leadership Turmoil, Cost Cuts Mount
• Inside AI’s biggest gathering, a surprising admission: No one knows how it works
• Sabre Corporation (SABR) Launches AI Chat Tool for Airlines
• Microsoft Is Reportedly Struggling to Sell Its AI Products
• Brook.ai Wins Top Digital Health Award for Pioneering Work in AI-Powered Remote Care Management
• Opinion | This Is the 21st-Century Arms Race. Can America Keep Up?
• Bill Gates on why AI will become 'hyper competitive'
• Rise of ‘rogue’ AI agents pushes companies to lock down data
• Google LiteRT NeuroPilot Stack Turns MediaTek Dimensity NPUs into First Class Targets for on Device LLMs