Microsoft Majors on AI with Both Software and Hardware
Unsurprisingly, much of the focus at Microsoft Build 2024 was on impressive new AI capabilities possible in Windows, which will all be boosted by a new Copilot+ PC range.
New Copilot+ PCs Launched
Ahead of Microsoft Build 2024, Microsoft held a non-streamed event that saw it launch a whole new range of laptops dubbed Copilot+ PC and designed specifically for AI.
Featuring new versions of the company’s own Surface Pro and Surface Laptop, and with around 20 models due to launch in mid-June from a stellar list of OEMs including Acer, ASUS, Dell, HP, Lenovo, and Samsung, the initial releases are based around Qualcomm’s new Snapdragon X Elite and Snapdragon X Plus chipsets.
Intel and AMD versions will follow in the future as Microsoft is keen for the moment to highlight its pivot to Arm-based processors. It is also drawing as much attention as possible to the performance advantage the Qualcomm chips have over current Apple silicon, quoting figures of 58% faster in sustained multithreaded performance over Apple’s M3-powered MacBook Air 15-inch. Impressively, this is claimed while still delivering all-day battery life, undoubtedly one of the holy grails of laptop development.
This is a huge endorsement for Qualcomm’s new processors as it looks to disrupt the PC world following similar moves in automotive and further diversify its base. Moving into the PC market was — and still is — a long term play for the company that has already required significant time and resources in an effort to steal market share from established players. Maintaining the pace of innovation will be paramount to ensure differentiation and stay ahead of rivals. As far as launches go, it is all looking positive so far.
It is worth pointing out that not all of the Copilot+ PC performance is solely chip-based. Microsoft has also re-architected Windows 11 to run multiple AI models as background tasks and make the most of the new silicon triptych of CPU, GPU, and the new generation of high-performance Neural Processing Units (NPU) designed for AI processing.
Pricing is surprisingly competitive, especially as badging OEM machines with Copilot+ PC branding requires a minimum spec of a 256GB SSD, an integrated NPU, and 16GB of RAM. Machines from the OEMs and Microsoft itself start at $999 for Snapdragon X Plus powered units (Snapdragon X Elite machines add $300), which is perhaps why Microsoft is forecasting that 50 million Copilot+ PCs will sell in their first year.
One of the reasons for that minimum spec is that the machines will need plenty of power to run some of the new on-device AI features that were announced at the event. These were headed by Recall, effectively an AI-powered search function that logs everything a user does by taking an encrypted image of the active screen every few seconds. This is fully searchable, Recall offering different options to interact with the surfaced content. Microsoft was at pains to point out that all analysis is done privately and on-device.
Other demos targeted creatives, with the promise of optimised experiences that leverage the power of the new Snapdragon on-board NPU in Adobe, Davinci Resolve Studio, CapCut, and more.
Make no mistake, these are powerful machines that take the game to Apple, and it will be interesting to note real world benchmarks when they are released. They mark the beginning of a new phase of the AI rollout that ironically harkens back to previous eras when hardware performance was at a premium. Privacy and latency issues mean that on-device processing will be crucial for the next wave of AI applications. Vendors will be hopeful that they can initiate a new supercycle of device replacements as a result.
The question is whether consumers will be convinced. The new machines are competitively priced at the mid to high-end of the laptop market that Apple has colonised. However, the software offering and the abilities the Copilot+ PC unlocks will have to be extremely compelling for people to want to short-circuit that upgrade cycle. A strong developer community is therefore going to be critical, which brings us to the news from Build 2024.
Microsoft Build 2024 in Focus
Once Build 2024 got started a day later, Microsoft CEO, Satya Nadella, kicked off a two-hour plus presentation that set out the company’s stall for the continuing rollout of what we might as well call AI Everywhere with a slew of announcements.
Nadella stated that Microsoft had harboured two dreams for decades: can computers understand us rather than us having to understand computers? And in a world of increasing information, can computers help us reason, plan, and act more effectively on that information?
Build started as a developers’ conference and has very much kept to that mission. A lot of its announcements are still very much targeted at the developer community rather than the wider audience that, say, Google I/O and the forthcoming Apple WWDC seek to engage with.
They still reverberate though. Microsoft’s Windows chief, Pavan Davuluri, argued that Windows is currently the most open platform for AI development and said that the company now has more than 40 AI models running on Windows 11. He highlighted the new Windows Copilot Runtime library of APIs that will allow developers to use these models for their apps. The hope is that simplifying the development path will lead to a greater library of AI-powered apps being created for end users.
The Copilot stack that Microsoft has developed is designed to help the company and the development ecosystem build AI tools faster. To that end, the new Microsoft Copilot Studio introduces new agent capabilities, letting developers build Copilots that can proactively respond to data and events, tailored to specific tasks and functions. The company states that Copilots built with this new category of capabilities can now independently manage complex, long-running business processes by leveraging memory and knowledge for context, reason through actions and inputs, learn based on user feedback, and ask for help when they encounter situations that they don’t know how to handle.
Use cases mentioned by the company ranged from IT device procurement to customer concierge for sales and service. And Copilot extensions, including plugins and connectors, allow customers to further enhance Microsoft Copilot by connecting it to new data sources and applications, expanding its functionality.
More consumer-facing, Microsoft also previewed Team Copilot, an upgrading of Copilot for Microsoft 365 from a behind-the-scenes, personal AI assistant to what it describes as “a new, valuable member of your team” in collaborative software such as Teams, Loop, Planner and more. Examples highlighted by the company here included working as a meeting facilitator in meetings, managing an agenda, tracking time and taking notes; surfacing important information, tracking action items, and addressing unresolved issues in chats; project management, and more.
There was more hardware announced too with the unveiling of a mini PC featuring the Snapdragon X Elite chip along with 32GB RAM and 512GB of NVMe storage. Currently, this $899 Snapdragon Dev Kit for Windows is not destined for general sale. This could change, however, if interest is high and Microsoft could find itself targeting the Mac mini market as well.
Elsewhere:
OpenAI’s newest flagship model, GPT-4o, is now available in Azure AI Studio and as an API
There was some emphasis on the cost savings (12x cheaper to make a call) and speed updates ( 6x faster in time to open response) GPT-4o enables
Microsoft and Cognition will bring Cognition’s autonomous AI software agent, Devin, to customers to help with complex tasks such as code migration and modernization projects
Phi-3 vision was announced, a variant of the company’s Phi-3 Small Language Model series designed for mobile devices, which can perform general visual reasoning tasks such as answering questions about a chart or specific images
Microsoft is enabling Khan Academy to offer all K-12 US educators free access to Khanmigo for Teachers, an AI-powered teaching assistant. Khan Academy is also collaborating with Microsoft to explore opportunities to improve math tutoring with a new version of Phi-3
Nvidia GPUs will ship in Copilot+ PCs in the next few months
Riding the AI Wave
All in all, this is a huge moment for Microsoft as it looks to position itself at the forefront of what Microsoft CTO and EVP of AI, Kevin Scott, called “an extraordinary platform wave, where something is fundamentally changing in the universe of technology.” He likened it to the PC revolution when Moore’s Law was driving forward a huge increase in power along with a concomitant lowering of the cost of personal computing.
Microsoft was one of the big winners of that epoch. Having famously then missed the start of the revolution in connectivity that was the early days of the internet, the message from this week is that it has no intention of being left behind again. It is not only trying to catch the wave, it is trying to own it.
It is worth reiterating that it is a huge moment for Qualcomm too. The company has gambled considerably on its entry into the laptop market, and there is still the hurdle of real-world testing to surmount. Regardless, though, the move has paid off given its initial high-profile association with the Copilot+ PC range as other chip manufacturers will only be part of it at a later date.
All eyes now turn to Apple and WWDC in June. That is 18 days away. If a week is thought to be a long time in politics, two and a half weeks seems like a geological age in AI. Apple’s plans will have been long in motion and it will still be no surprise to see it issue a counterpunch to all the latest news and try to reassert itself firmly at the forefront of consumers’ minds when it comes to AI.