Is everyone making chips now?

The Chip Race Is On

The race to develop powerful silicon chips for cutting-edge AI models is kicking into overdrive. In just the past week, we saw a slew of new custom chip releases from tech heavyweights.

Up first: Intel. On Tuesday, Intel showcased its new Gaudi 3 chip at its Vision event. Intel claims that Gaudi 3 is 50% faster than Nvidia’s H100, with up to 1.7x better performance on LLMs—at a lower price point.

Then came Google. On the same day, Google also took aim at Nvidia’s chip dominance by announcing the general availability of two new AI chips at Cloud Next:

1. Cloud TPU v5p, which can train LLMs almost 3x faster than previous TPU v4. These chips are used as alternatives to Nvidia’s GPUs for AI acceleration tasks.  

2. Axion CPU: Google's first Arm chip, promising 30% better performance and 60% better efficiency than the fastest general-purpose Arm-based instances available in the cloud today. These chips are used to support Google’s AI work in data centers.

And finally, Meta. On Wednesday, Meta unveiled the "next-gen" Meta Training and Inference Accelerator (MTIA) chips—promising “3x better performance” compared to MTIA v1. The specs:

• Built on 5nm process (vs. v1’s 7nm) with more cores

• More internal memory (128MB vs. 64MB) and faster speeds (1.35GHz up from 800MHz)

Meta said that MTIA “will be an important piece of its long-term roadmap” to build the best possible AI infrastructure.

So…what about Nvidia? While Intel has been a long-time Nvidia competitor, Google parent company Alphabet and Meta are two of Nvidia’s biggest customers. And though Nvidia currently holds a 75% share of the AI accelerator market, these three chip arrivals are already threatening its dominance—shares are down around 10% from recent highs.

Is Udio the Future of Music Generation?

This week, Suno competitor Udio hit the market—claiming sound quality that “meets professional standards.”

What is Udio? Udio is a new AI music generator founded by four former researchers from Google’s DeepMind with a mission of “making it easy for anyone to create emotionally resonant music in an instant.”

• And it’s backed by big names: The company raised $10 million in seed funding from investors including Andreessen Horowitz, Instagram’s co-founder Mike Krieger, and musicians Common and will.i.am. 

How it works: Once a user enters a text prompt (detailing genre, lyrics, and artists that inspire), the tool…

• Uses an LLM (like ChatGPT) to generate lyrics.

• Synthesizes music with a diffusion model like Stability AI’s Stable Audio. 

• Generates two song snippets for the user to choose from, which they can then download, remix, extend, and share on the app’s built-in community or social platforms.

What makes it different? If you’ve tried Suno, you’ve basically tried Udio. The main difference is that Udio (which was crafted with musicians in mind) offers users more intricate fine-tuning elements like instrument selection and sonic textures. If you want to see it in action, check out my full video unpacking Udio.

Optimize Your YouTube Content in One Click

Are you making the most of your YouTube content?

Your long-form YouTube content is a goldmine of potential Shorts—and if you aren’t making Shorts, you’re missing out on potential views, new subscribers, and audience engagement.

That’s why over 10 million creators and businesses use TubeBuddy’s AI-powered Suggested Shorts feature. This tool can:

• Analyze and identify the most engaging parts of your video.

• Recommend up to eight clips tailored to your specific content and audience.

• Get expert advice to enhance titles, descriptions, and tags.

The Next Wave: Your New Chief AI Officer

Looking for your Chief AI Officer? I’ve got you covered. This week, I launched my new podcast, The Next Wave, with my co-host (and incredible AI entrepreneur) Nathan Lands. 

In the first episode, we sat down with Aravind Srinivas, co-founder & CEO of Perplexity, the AI-chatbot-powered search engine (and one of my personal favorite AI tools). We cover a range of topics, including:

• Arivand’s journey from Open AI to Google DeepMind to Perplexity

• How Perplexity is differentiating from Google by “transitioning search results from links to answers”

• Open source vs. closed source models (and why Perplexity believes in open source)

• Apple is bringing AI-focused M4 chips to its entire Mac line.

• OpenAI fires two AI researchers for allegedly leaking information.

• Mistral releases its new Mixtral 8x22b model, one of the most powerful open-source AI models to date.

• Google expands its Gemma Family with models for developers and researchers.

• Cohere launches Rerank 3, a foundation model built to enhance RAG systems.

• Clean freak? You’ll love Dyson’s new AI feature that shows you exactly where you vacuumed.

• OpenAI updates its GPT-4 Turbo with Vision model, available via API.

More important AI news: Dive deeper into this week’s hottest AI news stories (because yes, there are even more) in my latest YouTube video:

The best AI music generator ever? Let’s try out Udio and Suno v3 together—and see where Udio might flop…