In front of Huang Renxun, Microsoft released its self-developed AI chip: running a large model, first meeting its own needs
ws_lmx
发表于 2023-11-16 10:10:53
3172
0
0
Technology giant Microsoft (Nasdaq: MSFT) has officially entered the AI chip industry.
On November 15th, Microsoft Chairman and CEO Satya Nadella delivered an almost hour long opening speech at the Microsoft Ignite Global Technology Conference. As predicted earlier, Nadella's speech almost entirely revolved around AI, launching the first CPU (central processing unit) Azure Cobalt 100 designed by Microsoft in terms of hardware, as well as the first AI chip Azure Maia 100 specifically designed for cloud training and reasoning. Both will be prioritized for supporting Microsoft's own cloud services.
On the same day, Microsoft's stock price remained stable, rising 0.04% to close at $369.67, continuing to reach a new closing high. Currently, Microsoft's total market value has reached $2.75 trillion.
According to Nadella, Cobalt 100 is based on the Arm architecture and has 128 cores. It is known as the "fastest CPU among all cloud computing providers" and has been applied in some of Microsoft's businesses and will be launched next year. The highly anticipated Microsoft self-developed AI chip Azure Maia also made its debut at the conference. The Maia 100 adopts a 5nm process and has 105 billion transistors. Nadella stated that Maia 100 aims to run large language models and help AI systems process large amounts of data faster. It will first provide support for Microsoft's own AI applications, and then be open to partners and customers.
According to foreign media analysis, Microsoft's Maia 100 and Cobalt 100 will respectively pose challenges to NVIDIA and Intel's leading positions in the AI chip and CPU industries. Faced with the challenge of insufficient supply of top AI chips, Microsoft has finally joined the self-developed ranks of other internet giants. Previously, Google had launched its self-developed AI Tensor Processing Unit (TPU) since 2016 and had reached its fifth generation in September of this year. Amazon Cloud Technology (AWS) announced the launch of its self-developed chip Training for training AI models in 2020.
At the conference, Nadella also took the opportunity to showcase the close cooperation between Microsoft and two chip giants NVIDIA and AMD. Azure cloud services are about to use the latest NVIDIA H200 chip, while on the other hand, the computing power provided by AMD's AI chip MI300X will also be available to some Microsoft users starting from the 15th.
NVIDIA CEO Huang Renxun also made an appearance at the conference, reviewing the comprehensive cooperation process between the two AI giants from hardware to software, conducting a wave of "commercial hype", and announcing that Microsoft users will be able to generate and deploy models using NVIDIA's AI workshop services on Azure cloud.
When asked about the future prospects of AI innovation, Huang Renxun stated that generative AI is "the most important platform transformation ever in the computer industry". Huang Renxun believes that the first wave of AI originated from a group of startup companies such as OpenAI, and now the entire industry has entered the second wave, which is the enterprise level AI driven by Microsoft Copilot. The third wave of AI will also be the largest, with the entire world's heavy industry being digitized and benefiting from generative AI.
Of course, Nadella also didn't forget to bring "good friend" OpenAI in his speech.
Microsoft Azure Cloud Services will provide users with the latest OpenAI products as soon as possible, including the GPT-4 Turbo and multimodal capabilities announced at last week's OpenAI Developer Conference, as well as fine-tuning based on GPT-4. In addition, OpenAI's custom GPT function GPTs will be integrated into Microsoft's Copilot system, helping Microsoft launch Copilot Studio for custom AI assistants.
However, due to the overlap of business areas, the competitive relationship between Microsoft and OpenAI has also been receiving attention. The Verge technology website pointed out that Microsoft took advantage of this conference to officially rename Bing AI as Copilot, perhaps to better compete with ChatGPT.
Undoubtedly, Microsoft's huge "circle of friends" has provided many conveniences for the company's various business development. At the conference, as another major announcement, Microsoft announced the official opening of the "one-stop" data processing platform Microsoft Fabric. The design concept of this platform is to consolidate all data assets of the enterprise and reshape the existing way in which the enterprise processes data. Fabric's AI assistant can also integrate with Microsoft's office ecosystem such as Office and Teams, providing functions such as "one click PPT generation", and connecting data channels with a variety of SaaS (software operations services) companies.
In addition, Microsoft also announced today the launch of the "MAAS (Models as a Service)" industry model, where users can directly call major APIs (application programming interfaces) through Microsoft's services, fine-tuning or deploying various open-source large models. Nadella announced that Microsoft has reached a partnership with major model developers such as Meta to launch well-known models such as Llama 2 as a service.
CandyLake.com 系信息发布平台,仅提供信息存储空间服务。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
猜你喜欢
- Apple lowers prices of various iPhone models in India
- NVIDIA skyrockets by 8%! Huang Renxun reassures the market: The rush to buy AI chips continues, and everyone is counting on us
- Baidu Shen Dou: Upgrade computing platform capability for 100000 card computing power cluster, Wenxin large model daily usage exceeds 700 million times
- Meta releases heavyweight new products: $299 Quest 3S headset, AR glasses prototype, multimodal AI model
- Xiaopeng increases its layout in Shanghai: the chip team size has exceeded 200 people
- Is winter really coming? Micron's financial report hits face, leading to a collective frenzy of global chip stocks
- Baidu World 2024 will be held on November 12th, and the daily average number of adjustments for the Wenxin large model has exceeded 700 million times
- 挑战Model Y 蔚来的品牌下沉“阳谋”
- Ford CEO tired of making 'boring' car models, personalized and electrified products become 'new favorites'
- Dialogue | Baidu Li Tao: The overlap between automotive intelligence and the wave of big models is a historical inevitability
-
9月が終わり、映画・テレビ業界が暗躍している。最近、愛奇芸の創始者でCEOのGONG宇氏は、映画・テレビ業界が長短の変化、AIの変化、中国映画・テレビドラマの海外進出の変化の3つの変化を経験していると発表した。 ...
- 寒郁轩良
- 前天 14:17
- 支持
- 反对
- 回复
- 收藏
-
ネットワーク状況監視サイトDownDetectorによると、オーディオストリーミングプラットフォームSpotifyは日曜日に約3時間にわたる障害を経験した後、正常に回復し、ピーク時には米国の4万人以上のユーザーに影響を与 ...
- hecgdge4
- 前天 09:33
- 支持
- 反对
- 回复
- 收藏
-
百済神州(688235.SH)はA株の有名な革新薬企業で、2017-2024年上半期、同社の各期の研究開発投資はA株の化学製薬会社(申万二級)の中で最も高く、研究開発費用は合計600億元を超えた。 しかし、百済神州にとって ...
- wylz8473
- 3 天前
- 支持
- 反对
- 回复
- 收藏
-
10月1日、極クリプトン自動車が発表したデータによると、今年第3四半期に新車が累計14万2900台納入され、前年同期比81%増となった。このうち、9月に新車を納入したのは2万13万人で、前年同期比77%、前月比18%増だっ ...
- 内托体头
- 昨天 16:17
- 支持
- 反对
- 回复
- 收藏