What a turnaround for nuclear power — and it’s all thanks to artificial intelligence.
This week two more Big Tech companies bet on nukes as Google and Amazon joined the party Microsoft started a few weeks ago. All that training of AI models takes a lot of power. And that’s despite, or because of, the fact that AI models are getting smaller and easier to use, as a raft of announcements this week made apparent.
AI is also exploding the data center, as all the talk at the Open Compute Project’s Global Summit this week was about accommodating AI workloads — such as Crusoe’s plans for building an AI megafactory in Texas — and the massive increase in power, heat and networking needs they require. Indeed, as a result, networking is the not-so-new but now critical bottleneck in the data center. Companies such as Lightmatter and xScape Photonics are betting big on the opportunity.
Paul Gillin lays out the very public Snowflake-Databricks clash in detail in the definitive story of the latest battle of the AI era. Ironically, given all the heat AI is producing, it’s about Apache Iceberg, the management layer that sits atop data files in cloud storage.
Google shook up its management team, in particular in search and ads, as well as Gemini AI. It looks like DeepMind CEO Demis Hassabis has finally consolidated his hold on all things AI at Google.
Down is up and up is down in the chip industry. In a hell-freezes-over moment, Intel and AMD decided to cooperate on standardizing the x86 chip instruction set. Meanwhile, TSMC and ASML issued widely different earnings results — chipmaker TSMC benefiting from roaring AI chip sales and equipment maker ASML suffering from a drop in sales to China and Samsung.
Next week’s events include UiPath Forward and Celonis’ Celosphere, and we’ll have news and interviews on theCUBE from both. The enterprise earnings season kicks off with IBM, Seagate and SAP, ahead of a much large number of reports the following two weeks from Alphabet, Microsoft, Amazon and many others.
TheCUBE Research analysts John Furrier and Dave Vellante discuss this and other news in more detail on this week’s theCUBE Pod, out later today on YouTube. And don’t miss Vellante’s weekly deep dive, Breaking Analysis, arriving this weekend for some lean-back reading. Meantime, check out this past week’s installment, zeroing in on Microsoft’s much-debated Azure revenue and what it means for the overall cloud market.
Here’s all of this week’s top news, newsy topics and analysis from SiliconANGLE and beyond:
AI and data: Smaller, simpler, faster… and more contentious
Headline news and analysis
Big-data dust-up: Why two AI giants are at war over who is more open
Accel says generative AI startups get 40% of VC investment in cloud
New models and services
Mistral introduces Ministral 3B and 8B AI computing models for phones and laptops
AWS and Databricks to deliver more affordable generative AI for joint customers
LatticeFlow releases framework for checking LLMs’ compliance with the EU AI Act
Google looks to spearhead AI adoption in the public sector
DataStax merges its data stack with Nvidia’s development tools to simplify AI development and fine-tuning
Adobe introduces new generative AI features for its creative applications
Domino Data Lab seeks to embed governance in AI development
Cognizant enhances Neuro AI platform for faster AI use case deployment
OutSystems integrates generative AI with low-code to accelerate and simplify app development
Big data startup UltiHash unveils advanced deduplication to boost AI storage efficiency
Money matters
Terray Therapeutics rakes in $120M for AI-powered small molecule drug development
AI startups Decagon and Neuron7 raise millions to transform customer service and repairs
AI observability firm Galileo raises $45M to improve AI model accuracy
Omnea raises $25M to streamline procurement with AI
French AI startup Gladia raises $16M and launches multilingual real-time transcription engine
Data science and AI development platform Zerve raises $7.6M
Simplismart raises $7M to help enterprises run their own AI models with rapid inference and full control
Amplitude buys Command AI to help companies integrate AI-assisted onboarding and product tours into software
There’s moreAI andbig data news on SiliconANGLE
Around the enterprise: Big Tech goes nuclear
Completing the cloud nuclear trifecta after Microsoft last month announced it will tap into a reopened Three Mile Island: Google to buy nuclear energy from small modular reactor startup Kairos Power and Amazon to invest in three nuclear energy projects But these are all going to take years to produce energy, so it’s not going to solve the data center energy crunch anytime soon.
AI explodes the data center
At the Open Compute Project Global Summit in San Jose, which focuses on new technologies and standards for data centers, the talk was all about how AI is changing everything in the data center, and creating new problems to solve:
* Does it seem hot in here?: If there’s one thing everyone’s freaking out about, it’s heat — specifically the heat generated by servers full of chips from Nvidia and other AI accelerator chipmakers. There were dozens of companies offering liquid-cooling solutions, considerably more than last year if I remember correctly. “You can no longer air-cool these servers,” said Saurabh Dighe, corporate VP for Microsoft Azure strategic planning and architecture. They need to be liquid-cooled.”
* But AI chips aren’t the only source of heat. Copper wire works fine in individual data center racks, but with AI forcing more racks to be connected to tackle training of huge AI models, optical interconnects are becoming more and more critical — and those interconnects currently also take a lot of power and produce a lot of heat. “We need to get the power down,” said Andy Bechtolsheim, co-founder and chief architect of Arista Networks. “We cannot change the laws of physics, but we can come up with better connectors.” That’s something established companies and startups such as Lightspeed are working hard to do.
* Sustainability was the other, closely related byword at the summit. One standard being mulled is how to use “all that warm water from [cooling] data centers for more than just letting it evaporate in the air,” said Supermicro VP of Solutions Thomas Garvens.
* But the problem with standards is that they take a lot of meetings and back-and-forth to establish. “Standards can take more time than we have in the era of AI,” said Google Principal Engineer Amber Huffman.
* By the way, if it weren’t apparent already, networking looks to be the biggest bottleneck in the AI era. “The networking is essentially becoming the computer,” said Sameh Boujelgene of market researcher Dell’Oro Group.
* ”Pigs can fly”: AMD EVP Forrest Norrod used that phrase to describe a deal between AMD and Intel to create an x86 Ecosystem Advisory Group, which although not an OCP project was revealed onstage at the summit by Norrod and Intel EVP Justin Hotard. If Intel weren’t in such a hole, this never would have happened, given the decades of legal and market battles between the two. But if either of them wants to challenge Arm, whose designs are now powering data center chips for cloud providers such as Amazon Web Services, Google and Microsoft, they need an instruction set as unified as Arm’s to ensure software compatibility among Intel and AMD chips and keep developers on the x86 train. Our story: Intel and AMD team up to improve the x86 processor architecture
There was lots of other news, honestly a lot of pretty arcane stuff, but no surprise that Nvidia, whose CEO Jensen Huang insists all data centers must become “AI factories,” captured so much attention: Nvidia aims to boost Blackwell GPUs by donating platform design to the Open Compute Project
And this startup is betting big on one ginormous factory: Crusoe’s $3.4B joint venture to build AI data center campus with up to 100,000 GPUs
Money matters
Breaking Analysis: What Microsoft’s financial disclosures reveal about Azure’s market position And it may force other big companies to reveal more about their AI revenue.
Private equity is bottom-fishing for software-as-a-service companies, which are increasingly being pressured by generative AI — not just because of investor interest but because chatbots have the potential to replace a good chunk of SaaS: Permira completes its $7.2B acquisition of Squarespace and Zuora to be acquired by Silverlake and GIC in $1.7B cash deal
Lightmatter raises $400M at $4.4B valuation for next-gen photonic data center networking
Xscape Photonics raises $44M to accelerate development of laser-based chip interconnects
ASML shares plunge on lowered 2025 revenue forecast
TSMC shares jump on expectation-topping third-quarter results
Dialpad expands workforce management capabilities with Surfboard acquisition
Developer tooling startup Port sails away with $35M in funding
Everstage raises $30M to enhance sales performance management platform
Lidar chip startup Lidwave closes $10M investment
Clerk Chat raises $7M to simplify business text messaging
VMware settles securities fraud class suit with $102.5M payout (per The Register)
New services
Salesforce launches customized financial services cloud platform for insurers
HashiCorp debuts new Terraform Stacks and Vault updates at HashiConf 2024
Tableau Software updates data visualization and cloud analytics management
SUSE enhances edge computing with new releases focused on scalability and efficiency
Cato Networks expands SASE Cloud Platform with new Digital Experience Monitoring service
Presidio expands AWS partnership to drive AI adoption and cloud migration
We have more news oncloud,infrastructure andapps
Cyber beat: Securing enterprise Android
New services
Google enhances security and management for corporate-owned Android devices
Akamai introduces Behavioral DDoS Engine and AI Assistance to strengthen security portfolio
Bitdefender’s new AI copilot works in real time to protect consumers from online scams
Lookout adds ‘smishing’ and executive impersonation protection to mobile security platform
Nametag launches ‘Deepfake Defense’ and expands Okta partnership to combat AI impersonation
OneSpan introduces VISION FX for phishing-resistant banking security
Tenable brings intelligent data and AI resource security risk management to the cloud
Money matters
Cyera acquires Trail Security for $162M to expand AI-powered data security platform
DeNexus expands cyber risk management platform with $17.5M investment
Morecybersecurity news here
Elsewhere in tech: Kindle gets color
Amazon refreshes its Kindle e-reader lineup with four new devices
FTC makes canceling subscriptions less of a pain with ‘click to cancel’ rule
Google files for emergency stay against Epic lawsuit it lost
Sam Altman’s Worldcoin rebrands as World and launches updated iris-scanning Orb
Stripe reportedly in talks to acquire fintech startup Bridge for $1B
Blockstream raises $210M to boost bitcoin blockchain layer 2 scaling technology
Path Robotics secures $100M Series D funding for its AI-enabled robotic welding technology
IoT startup Monogoto raises $27M to accelerate its global expansion
And check out more news onemerging tech,blockchain and crypto andpolicy
Comings and goings: Google’s management shakeup
Google reshuffles product leadership team and two AI units DeepMind CEO Demis Hassabis gets the rest of Gemini, cementing his hold on all things AI at Google and elevating him to perhaps the No. 2 exec under Alphabet and Google CEO Sundar Pichai.
Meta announces layoffs across several divisions for abusing meal credits
Sebastien Bubeck, Microsoft’s vice president of generative AI research, is leaving the company to join ChatGPT maker OpenAI, though it’s not clear what role he will assume (per Reuters).
Former Palantir CISO Dane Stuckey also joins OpenAI to lead security (per TechCrunch)
What’s next
Events
Oct. 22-23: UiPath Forward, Las Vegas: TheCUBE will be there.
Oct. 23-24: Celonis’ Celosphere, Munich: TheCUBE will be there.
Earnings
A couple of big enterprise reports to kick off the third-quarter earnings season:
Monday, Oct. 21: SAP
Tuesday, Oct. 22: Seagate
Wednesday, Oct. 23: IBM