HARDWARE

How Qualcomm just shook up Nvidia in AI chips

Qualcomm may have the open-platform answer to disrupt Nvidia's death grip on the AI ecosystem. 

The chip company, best known for its work in mobile devices, made a series of announcements over the past week that make it clear it's plowing full steam into AI, data centers, and playing a role in alleviating the compute crisis caused by AI. 

"We're now in this transition to what we're going to do in the next five years," Qualcomm CEO Cristiano Amon told investors on June 24 at Qualcomm Investor Day in New York City. 

Qualcomm's big moves include:

• Launching Dragonfly as its AI data center brand: Qualcomm now has the Dragonfly C1000 CPU, a Dragonfly AI300 inference accelerator, and a new memory technology called High Bandwidth Compute (HBC). These are especially designed to run AI workloads efficiently, saving power and reducing costs. 

• Acquiring Modular to create an open platform: Modular lets developers write AI software once and run it on any type of chip. That will now include Qualcomm chips, as well as potentially AMD, Intel, and others. This will free AI builders from Nvidia's lock-in. 

• Signing Meta as its first customer: Meta remains a big AI spender and signed a multi-generation deal in which Qualcomm's Dragonfly C1000 CPU will power Meta's next-generation servers, with production starting in the second half of 2028.

• A Hugging Face deal for deploying AI: Qualcomm signed a deal with Hugging Face to enable its 16 million developers to run its giant library of over 3 million open AI models on Qualcomm chips, from phones to laptops to cars to AI data centers.

"From a high-performance compute solution standpoint, we have four different kinds of solutions that actually feed into the data center business," Qualcomm's GM of data center solutions Durga Malladi told The Deep View, referring to XPUs, custom CPUs, HBC memory, and Alphawave for networking.

Although the least flashy thing that Qualcomm announced was the Modular acquisition, it's potentially the most significant. With Modular's software, Qualcomm can create an open platform that allows companies to run a "multi-silicon token factory," as Qualcomm called it during its Investor Day presentation. 

For many AI businesses, that could provide a degree of freedom from Nvidia, as Modular is a more modern competitor to Nvidia's CUDA software, which most of today's AI models use to access GPUs. Modular's software can access Nvidia GPUs faster and can access processors from other chipmakers as well.

"CUDA and these 20-year-old C++ technologies are gatekeeping GPUs," Chris Lattner, CEO of Modular, told The Deep View in March at Nvidia GTC. "Our software runs on CPUs, it runs on GPUs, and it can run ASICs, and you can have a consistent experience across all of them… We have the first software stack that can actually scale across heterogeneous compute."

Nvidia is estimated to hold about 80% to 90% of the market for AI GPUs used to train and run AI models. As much as AMD, Intel, Amazon's Tranium, Google's TPUs, Cerberus, and others have tried to compete with Nvidia, the company continues to dominate the landscape for powering frontier models. Qualcomm jumping into the fray on AI chips could ease some of the demand and put downward pressure on prices by using Modular at the software layer, enabling companies to mix and match hardware from various chipmakers. That's a good sign for the long-term potential to drive down the cost of AI. And don't worry too much about Nvidia. It reportedly has a 12-month backlog of orders for its chips. And it's been expanding aggressively into other technologies such as robots, self-driving cars, world models, open models, and digital twins. It's well aware that the CUDA lock-in won't last indefinitely. 

TOGETHER WITH MERCURY

Banking* that does financial work for you

You've seen what AI can do when it's built into the right system. Mercury Command is that for your finances.

Built directly into your Mercury account, Command turns plain language into executed financial work — categorize transactions, adjust a card limit, send an invoice — with full account context, explicit approvals, and a traceable record behind every action. 

Try Mercury Command →

POLICY

The politics behind Anthropic's California deal

California is getting a big discount on its AI bills. 

On Monday, Anthropic and California Governor Gavin Newsom struck a deal to give state agencies expanded access to the company's Claude family of models at half the cost, according to a Politico report. The partnership applies to both state agencies and local and city governments and includes free workforce AI training and technical support.  

Chris Given, the state's CIO and director of the state department of technology, told Politico that the intent is to get departments to switch their usage to this contract to save on costs. “When we see that folks are going to be using a tool more, we want to make sure that we, as the state, have negotiated the best possible price for them.”

The report did not specify whether this includes any access to Anthropic's Mythos model following the US government's decision that it could be deployed to a small subset of US organizations for defending critical infrastructure. According to Politico, Claude is already in use by the state for certain tasks, including patching code.  

The decision comes as California seeks to reckon with AI's potential to disrupt the economy and labor market. In late May, Newsom issued an executive order directing the state to prepare workers for AI's economic impact, enlisting state agencies, labor experts, economists, universities, and AI leaders to develop policies and monitor for warning signs. It's the latest in a string of state moves aimed at preparing for AI disruption. It's taking on issues such as safety and privacy long before other states or the federal government. 

"California has never sat back and watched as the future happened to us – and we won’t start now," Newsom said in a statement about the executive order. 

Additionally, the move echoes a playbook used by Anthropic (and every other major AI firm) last fall, when companies including Meta, Microsoft, OpenAI, Amazon Web Services, Oracle and others offered their services at steep discounts for periods of one to two years.

The partnership is ideal for both Anthropic and California. Given that Anthropic has built its entire brand around being the safest and most trustworthy AI lab, the company's models are a natural choice for a regulatory-forward state like California. And it isn't bad PR for either entity involved. But Anthropic stands to gain a lot more than just another contract and good press from its partnership with the state. For one, by integrating into state government systems, the partnership naturally lends itself to consideration for more government contracts, opening the door to another potential source of significant long-term revenue. Additionally, given that California tends to be on the leading edge of tech regulation in the US, Anthropic partnering with the state could give it an avenue to influence a regulatory environment that is far from set in stone. 

TOGETHER WITH MODE MOBILE

Apple Price Hikes Reveal Mode’s Big Opportunity

Apple will potentially raise prices on Macs and iPads by $200 or more, another reminder that the devices we rely on keep getting more expensive.

Mode Mobile, is building the countertrend: technology that turns everyday smartphones into income-generating assets.

Its EarnOS platform has already helped users earn and save $1B+, reached 490M+ users, and powered 32,481% three-year revenue growth, earning Deloitte’s #1 fastest-growing software company ranking.

Now, Mode has secured the Nasdaq ticker $MODE, and its pre-IPO investment offering is open at $0.52/share with up to 20% bonus shares. But that window is closing fast.

Click here to learn more.

WORKFORCE

Ford discovers AI can't replace deeper expertise

Another company is regretting its decision to lay off employees in favor of AI. 

Automotive giant Ford has reportedly hired around 350 veteran engineers, several of whom are former employees, after its AI systems couldn't deliver the results the company anticipated, Bloomberg reported.

Ford COO Kumar Galhotra told Bloomberg that the company had been increasingly relying on "automated quality systems,” and brought back technical specialists to find failure points "before a part ever reaches the plant floor." 

Additionally, Charles Poon, the automaker's VP of vehicle hardware engineering, said that the company "mistakenly" thought that introducing AI that ingested design requirements would result in a high-quality product. "We recognized that for us to enhance some of our automation and machine learning and artificial intelligence tools, we needed to ensure that they were trained by the most experienced individuals," he said.

Though Ford is far from abandoning its AI plans, it provides an example of what can happen when tech companies jump the gun on using AI to automate employee work. It's also just the latest signal that AI may be giving the job market some whiplash. 

A report from PwC published in mid-June found that companies using AI to augment and supercharge their workforces are seeing more returns than those that simply replace large chunks of their staff with automation. A report from Gartner published earlier this year predicted that more than 50% of the companies that cut staff due to AI, aiming to save on payroll, would end up rehiring employees to perform similar work.

Though there's a lot of talk about how AI will change the nature of work and create new jobs, Ford's decision to bring back staff is a sign that the tech may simply not be as prepared as we anticipated to handle certain tasks. It's why the recipe for success isn't worker replacement, but rather worker augmentation. Both the employees and the AI systems get better working in tandem. Then, when the systems are ready, the employee can take on more meaningful work. Additionally, not all jobs are equally ready for integrating AI. Physical AI systems for blue-collar work have further to go than AI for white-collar work. And, given the potentially dangerous nature of industrial and manufacturing settings, the stakes of getting it right are high. 

LINKS

• OpenAI, Booz Allen partner to advance AI for national security, defense

• Google connects Personal Intelligence to Nano Banana, Google Photos

• HP inks partnership with OpenAI to deploy Frontier platform

• Tidal to automatically restrict AI-generated music, blocks royalties

• Waymo, Uber quietly broke up partnership in Phoenix in May

• Meta reportedly places strict internal usage limits on Claude, Codex

• Brain2Qwerty: Meta has released the full training code for its end-to-end pipeline capable of real-time sentence decoding from raw brain signals 

• Gamma: The presentation generation and editing platform is now available in ChatGPT. 

• Scout: A new agent by Giga, keeping track of your KPIs, learning from conversations and improving your results.

• Claude: Anthropic's flagship AI models, Claude Opus 4.8 and Claude Haiku 4.5, are now available in Microsoft Foundry.

• Amazon: AI Builder, Ring

• Sesame: ML Model Serving Engineer

• OpenAI: Economist

• Nvidia: Senior Scientist, Synthetic Data Generation

GAMES

A QUICK POLL BEFORE YOU GO