“The Age of Copilots is Upon Us”
Or so Satya wants us to believe and buy into the vision!
Microsoft has missed out on many key waves of technology most notably getting mobile wrong. But this time, with Satya at the helm things are different. After getting cloud right, Microsoft’s going all in on AI – and the Ignite 2023 event was a good indicator of this.
Microsoft’s AI strategy seems quite straightforward:
- Look at the entire AI stack as shown above
- On each layer – Partner with market leaders or build it in-house and offer it as a service on Azure
- Deeply integrate each layer into the rest of the Microsoft ecosystem
- Name products better this time 😆
So let us begin with the base layer of infrastructure, the layer with the highest competitive moat and led by the rocketship that is Nvidia🚀
1. AI Infrastructure
- Azure Boost – Offloads server virtualization processes onto purpose-built software and hardware enabling massive improvements in networking speeds. Useful, also locking you further into the Azure ecosystem 😉
- Azure Cobalt CPU and Maia AI Chips – Late, but Microsoft has finally arrived into the custom silicon game. After the surge of demand for Nvidia’s H100 chips, every cloud provider has realized that they cannot be dependent on an external company for such a key component driving AI. Google had announced its own custom silicon TPU v5e at Google Cloud Next a couple of months ago and AWS is already ahead with its range of obviously-named custom chips – Graviton, Trainium and Inferentia. Some marketing agency charged thousands for these names 😒
- Why custom chips? – Eventually the LLM game will be dominated by a handful of large companies with the largest datasets available (or crawled 😉). The true value of AI will be found with enterprises and individual developers who will build customized use-cases on top of the base models. And their selection of the cloud platform to power their AI workloads will boil down to cost and the software tools to ease development. So while the cloud companies will insist they aren’t competing with Nvidia or AMD and continue to offer their chips, they will continue to develop their own chips optimized for the workloads on their platforms, driving down costs while also reducing their dependence on the chip companies and their proprietary tools and standards. Oh, and they literally said so and put themselves alongside the chip companies
2. Foundation Models – Azure AI
Onto the next layer which is quite straightforward. Azure wants to be the cloud of choice for ALL the AI models you’ll ever need as a service – GPT-4 Turbo, DALL E 3, Stable Diffusion, Mistral 7B, Nvidia Nemotron-3, Llama 2, Jais, Phi-2 etc etc and Azure AI Studio for all the tools you need to train, evaluate and deploy whatever AI model your heart desired. Oh and Nvidia founder and CEO Jensen Huang joined Satya to announce their own industry-focused solutions such as Omniverse will also be available on Azure.
3.Data Layer – Microsoft Fabric
Finally someone in Microsoft realized hey we have so many discrete data connection, analytics and visualization tools from PowerBI to Synapse – why don’t we combine this into one SaaSified offering? Interesting if they continue to build this out, someone in the Databricks team should start to get worried.
4.App Layer – Adding AI to Everything
- Teams at this point is just Microsoft’s AI test bench – they are pouring everything into it and every component from video streams to audio processing have tons of AI enhancements. At some point, Teams could become almost a chromebook like lite OS for enterprises, Windows will become redundant for basic workflows if Teams itself can handle everything – from documents to collaboration to communication all supported with AI. If this happens, remember I called it first 😆
- Mesh was announced – which is just another metaverse wannabe with the same keywords avatar, immersive space, 3D blah blah. Doomed to fail
5. Microsoft Copilot – One Ring to Rule them All
- Satya essentially said in his polished, corporate-speak that hey trust us to build your all powerful AI assistant which will be integrated with everything you do, every Microsoft tool, every click on your browser, every document, spreadsheet, email or line of code your company will write and it’ll be useful. Well Microsoft has been THE enterprise productivity partner for large companies globally – and if they get Copilot right, it could revolutionize work as we know it. Can’t wait to have my AI avatar join meetings on my behalf and flash the product roadmap again and again 😅
There were cursory mentions of AI for elements, drug discovery etc, but Google is far ahead in leverage AI for fundamental research. For example DeepMind recently publishing a paper about a tool called graphical networks for material exploration (GNoME) which casually has predicted 2.2 million new materials, of which more than 700 have gone on to be created in the lab and are now being tested. Oh these are totally new to science, we only know about 48000 crystals in our entire scientific body of knowledge so no biggie. AlphaFold is doing similar cutting-edge breakthroughs for biological research and drug discovery – very interesting contrarian bets by the two companies.
Microsoft truly is focusing on the enterprise and workplace as its AI battlefields, and is positioned to dominate. Like most Youtube comments on his keynote, Microsoft has a visionary CEO in Satya who knows his tech and is getting it right again and again and again.
Leave a comment with your feedback and thoughts! ❤️
