• Everyday AI
  • Posts
  • Boosting Performance: Azure's Proprietary Data Center Chips Unveiled

Boosting Performance: Azure's Proprietary Data Center Chips Unveiled

ChatGPT o1 model reclaims top spot, ChatGPT voice model available on web, Atman’s Rain AI chip gains investments, Meta unveils new AI product team and more!

Author

Everyday AI
November 20, 2024

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Microsoft just unveiled some new AI processors at its Ignite conference. We sat down with the Senior Director of Azure Infrastructure to break it all down. Give it a listen.

🕵️‍♂️ Fresh Finds: Gemini’s API receives OpenAI compatibility, Meta’s $5B AI data center and a new reasoning model. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: ChatGPT o1 model reclaims top spot, ChatGPT voice model available on web, Atman’s Rain AI chip gains investments and Meta unveils new AI product team. For that and more, read on for Byte Sized News.

🚀 AI In 5: Is Ideogram 2.0 a new competitor to Midjourney? We dive in to compare. TBH, we were kinda impressed. See it here

🧠 Learn & Leveraging AI: Microsoft is making moves to stay competitive in the AI processing space. We break down exactly how they’re doing it. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about Microsoft Ignite, NVIDIA x Microsoft AI integrations, Microsoft's Azure AI Foundry, and OpenAI facing a new lawsuit. Check it here!

Boosting Performance: Azure's Proprietary Data Center Chips Unveiled ⚙️

Microsoft unveiled two new custom processors, Azure Integrated HSM for security and Azure Boost DPU for data processing, at their Ignite conference to enhance security and efficiency in Azure data centers.

This move helps to position themselves against competitors like NVIDIA and AMD.

We sit down with Alistair Speirs, Senior Director, Azure Infrastructure at Microsoft to discuss.

Join the conversation and ask Jordan and Alistair questions on Microsoft AI here.

Also on the pod today:

• Detailed explanation about DPU 🤔
• Role of the HSM in encryption 📟
• Chip specialization in the tech industry 🧑‍💻️

It’ll be worth your 21 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Postiz is an AI social media scheduling tool, Adobe Podcast helps your audio sound professional and Dokko is a conversational AI platform for knowledge sharing.

Anthropic – An updated Claude 3.5 Sonnet underwent the first-ever joint pre-deployment evaluation by the U.S. and U.K. AI safety bodies.

Salesforce – Salesforce has introduced Agentforce Testing Center.

Business of AI – Cision is collaborating with Google Cloud to transform AI-powered communications.

Meta – Meta is planning a $5 billion AI data center in Louisiana

Google – Gemini’s API has now received OpenAI compatibility.

AI Models - Niantic, creator of Pokemon Go, is building a ‘geospatial’ AI model based on Pokémon Go player data.

This new reasoning model is apparently giving OpenAI’s o1 a run for its money.

1. ChatGPT’s o1 Model Levels Up and Regains Top Spot 👑

In a timely update, OpenAI has unleashed enhancements to its GPT-4o model, reaffirming its dominance in the language model arena. This upgrade boosts the model's creative writing capabilities, making it more natural and engaging. As highlighted by OpenAI, these advancements promise improved relevance and readability, suggesting that users can expect a more tailored experience when generating text.

2. ChatGPT's Voice Mode Hits the Web 🗣

OpenAI has just announced that its Advanced Voice Mode feature is now available for ChatGPT users on the web, expanding from its earlier mobile app release. This exciting update allows subscribers to engage in more natural, real-time conversations with the AI, complete with emotional responses and the ability to interpret non-verbal cues.

Users can choose from nine distinct voice options, tailoring their interaction to suit their preferences.

3. Altman Rallies for Rain AI in Chip Challenge 💰

Sam Altman is actively courting investors for a $150 million Series B funding round aimed at bolstering Rain AI, a promising contender in the semiconductor space, valued at approximately $600 million. This move comes as NVIDIA, which dominates 85% of the AI chip market, strengthens its ties with Elon Musk's xAI, raising concerns for Altman about potential competition.

With Rain AI touting chips that are more energy-efficient and powerful, Altman's backing could shine a spotlight on the startup, potentially reshaping the landscape of AI hardware.

4. Meta Unveils New AI Product Unit for Businesses 👥

Meta is launching a new product unit focused on developing AI tools tailored for the 200 million businesses that utilize its platforms, as reported by Axios. Clara Shih, formerly CEO of Salesforce AI, will spearhead this initiative, which aims to enhance customer interactions through innovative products like a customizable AI chatbot for WhatsApp and Messenger.

With Meta's AI chatbot already amassing over 500 monthly active users since its April rollout, the company is poised to deepen customer relationships beyond mere ad optimization.

5. OpenAI Launches Free AI Training for Teachers 🧑‍🏫️

OpenAI and Common Sense Media have unveiled a free training course designed to simplify artificial intelligence and prompt engineering for teachers. This initiative comes as ChatGPT continues to gain traction in classrooms, with many educators now exploring its potential beyond concerns of cheating.

The course aims to empower K-12 teachers to leverage AI for lesson planning and streamlining administrative tasks, reflecting a growing recognition of AI skills as vital for future careers.

6. HarperCollins Enters AI Training Deal with Microsoft 🤝

HarperCollins has announced a three-year agreement to allow Microsoft to use its content for training a forthcoming AI model, as reported by Bloomberg. This deal, which will focus on select nonfiction backlist titles, requires authors to opt in, with one author revealing an offer of $2,500 per book for participation.

Interestingly, Microsoft reportedly has no plans to produce AI-generated books from this material, raising questions about the intended applications of the AI.

7. Deloitte Partners with HPE and NVIDIA to Boost AI Adoption 🚀

Deloitte is teaming up with Hewlett Packard Enterprise (HPE) and NVIDIA to launch HPE Private Cloud AI, a platform aimed at accelerating corporate adoption of artificial intelligence. The initiative combines HPE's cloud and data storage capabilities with NVIDIA's AI infrastructure to deliver tailored on-site and co-location solutions, addressing scalability challenges many companies face.

Despite this promising collaboration, both HPE and NVIDIA stocks saw a slight dip in early trading on Wednesday, highlighting the cautious market sentiment.

Ideogram 2.0 Review: Does It Compete with Midjourney?

Is there a new Midjourney competitor on the scene?

Ideogram’s 2.0 update is pretty dang good.

Does it compete with Midjourney? We take a look.

🦾How You Can Leverage:

Microsoft Just Made AI chips like 400% more efficient.

 (And somehow, that’s not even the biggest news.)

The trillion-dollar war for AI infrastructure just got a plot twist: custom silicon that demolishes the "pick two" triangle of speed, cost, and security.

Cuz let's be honest, shorties.

Microsoft's new DPU chip just made every data center on Earth look ancient. (That’s a Data Processing Unit chip, for the non-technies among us.) 

 A single chip running at 100 watts now handles what used to require 400 watts of server hardware.

Dork talk? 

Maybe. But what Microsoft just unveiled at its Microsoft Ignite conference is going to profoundly change the technical side of cloud computing and how we all access AI. 

Alistair Speirs, Senior Director of Azure Infrastructure at Microsoft, joined the Everyday AI show to break down everything new for Azure at Microsoft Ignite.

But this isn't just about chips. Microsoft is building quantum-resistant data centers out of wood. Cross-laminated timber, to be exact.

(Yes, really.) 

They've also locked down 34 gigawatts of renewable energy through 2030. That's enough to power three New York Cities.

Whoa. 

Your current data center strategy is now changing, thanks to some new updates from Microsoft. 

What’s it all really mean? 

If you’re on the technical side of an enterprise, today’s show can be a lifeline. But even if you’re not, we’re about to get a whole lot smarter on why DPUs are so important.

Let’s break it down. 

1 – The 100-Watt Wonder: Azure's DPU Rewrites Physics

Traditional servers waste power like a broken faucet wastes water. Most CPUs multitask poorly, juggling storage operations while trying to handle everything else.

Microsoft's solution? Kill the multitasking.

Their new Application Specific Integrated Circuit does one thing: storage operations. Nothing else.

The result shatters efficiency records. A single DPU chip handles all storage tasks at 25% of the power draw.

By 2025, 40% of data centers will face power constraints. This chip just turned that crisis into an opportunity.

Try This:

  1. Document your current server power draw for storage operations

  2. Identify your top 3 storage-intensive workloads

  3. Calculate your actual cost per storage operation using Azure's pricing calculator

2 – Hardware Security Modules Get an Adrenaline Shot

Microsoft just eliminated the encryption tax. Their new HSM chip makes quantum-resistant security free from performance penalties.

Every server gets one. Not just special appliances tucked away in secure rooms.

The magic? Ephemeral encryption.

Keys exist for milliseconds, then vanish. Even processes on the same server talk through encrypted channels that disappear instantly.

This isn't incremental improvement. It's security without compromise.

Try This:

  1. List every system-to-system communication in your infrastructure

  2. Flag the ones skipping encryption due to performance concerns

  3. Create your transition plan for pervasive encryption

3 – Wooden Data Centers That Would Make a Lumberjack Proud

Steel and concrete are out. Cross-laminated timber is in.

These wooden data centers aren't just green marketing. They're engineering genius. Lighter frames mean less concrete foundation. Less equipment for construction. Lower carbon footprint.

But here's the real innovation: These centers "sip" and "slurp" power based on renewable energy availability. Your workloads automatically shift to match when clean energy peaks.

Microsoft's already secured renewable power contracts past 2030. They're not playing quarterly earnings games. They're reshaping infrastructure for decades.

Try This:

  1. Map your workload energy patterns against your local grid's renewable energy peaks

  2. Identify which processes can shift to match clean energy availability

  3. Build a 12-month transition plan for flexible workload scheduling

⌚

Numbers to watch

$40 Million

Federato, an AI startup that fixes insurance risk analysis with AI, has raised $40M

What are your thoughts on our Numbers to Watch section?

Login or Subscribe to participate in polls.

Reply

or to participate.