Every week NAND Research puts out a newsletter for our industry customers taking a look at what’s driving the week, and what happened last week that caught our attention. Below is a excerpt from this week’s, March 3, 2025.
Driving the Week
All eyes are turned towards Barcelona where the 2025 Mobile World Congress is happening this week. We’ve been pre-briefed on a lot of what’s being announced that’s relevant to enterprise IT: expect more edge, private 5G, and AI-enabled ORAN solutions. We have an analyst on-the-ground at MWC who will provide perspective as things are rolled out.
We’re in the heart of earnings season with plenty to watch for this week. Tuesday sees Crowdstrike releasing. Marvell and MongoDB announce on Wednesday. Thursday we’ll hear from Broadcom, HPE, and Elastic.
Policy Watch
Tariff Impact: computer-maker Acer is reportedly mulling a 10% across-the-board price increase to offset US tariffs against China. The company is also looking at moving some manufacturing away from China, but that’s a move that takes time – and they’d need to find a location that’s more tariff friendly to the US.
Export Controls: the US isn’t just controlling exports of AI accelerators to China, it’s also tightened access to countries like India, Switzerland, and Isreal. Microsoft is calling for an ease in restrictions in some of the “tier 3” countries to keep those countries from turning to China for advanced parts. NVIDIA agrees, previously calling the chip restrictions a “sweeping overreach.”
Legislating away unwanted content is challenging, but necessary, especially in this age of Deep Fakes. Last week the US Senate unanimously passed the Take it Down act which would make it a crime to publish AI-generated non-consensual “intimate imagery” while also calling for on-line platforms to implement reporting mechanisms to make it easy.
These things are never easy. Free-speech advocates, including the Electronic Frontier Foundation, immediately raised concerns about the free-speech implications, along with fears that the government could use the act as a stepping-stone towards more censorship efforts.
Enterprise AI
Maintaining its dominance, NVIDIA reported a staggering 78% year-over-year gain with its $39.3 billion top-line revenue for calendar Q4 2024. NVIDIA attributes much of the gain to “robust demand” for its Blackwell AI chips, noting that Blackwell is the fastest product ramp in the company’s history.
Looking beyond its open-source LLM, tech giant Meta is kicking off its new LlamaCon, a new developer conference dedicated to generative AI built around Meta’s popular Llama LLM and its associated Llama stack tools. The announcement came just as multiple outlets began reporting that Meta plans to release a standalone AI app as early as Q2.
Llama’s not the only open-source model gaining traction, as IBM releases the latest iteration of its Granite LLM. This is a nice release, incorporating conditional reasoning, time-series forecasting, and document vision capabilities. IBM continues to bet big on enterprise AI, and this is their year to shine.
But are there enough GPUs? This seems to be the perennial question, with OpenAI CEO Sam Altman recently raising alarms about the company’s dwindling GPU resources, stating that they are effectively “out of GPUs.”
Infrastructure
Storage is getting a whole lot faster: chipmaker Broadcom introduced its first PCIe 6.0 switch, doubling bandwidth and lowering latency. This is a critical piece in the puzzle for GenAI and real-time analytics workloads, and will no doubt drive a slew of storage controller upgrades across the storage industry – sometime next year.
Continuing the blur the lines between storage and data management, VAST Data is expanding its Datastore platform, adding block storage and event streaming capabilities, making it more adaptable to diverse workloads. We like where VAST is taking things, but we’re also still waiting to see market validation of its approach.
Like chocolate and peanut butter, observability and AI simply go together. Observability upstart New Relic is betting big on AI with new tools that analyze video quality of experience and user engagement. AI-powered observability is a natural for media, gaming, and streaming platforms, but it’s not an area we see many of New Relic’s competitors embracing. This is a nice move for the company.
The network world can be a confusing space. A prime example is the news that sometimes-competitors Cisco and NVIDIA are working together to optimize AI networking. While Cisco traditionally competed in data center networking, it’s now embracing NVIDIA’s dominance in AI infrastructurerather than fighting it. Expect AI-powered networking stacks that integrate NVIDIA’s DOCA software and BlueField DPUs.
Deal Watch
Doubling down on AI-powered search, database player MongoDB announced a $220 million acquisition of Voyage AI. The deal puts Mongo directly up against vector database specialists like Pinecone and Weaviate with its enhanced RAG capabilities. We expect to lead to MongoDB Atlas integrating smarter semantic search and AI-enhanced document queries.
Just as it finalizes its acquisition of HashiCorp, news arrives of IBM’s intent to acquire long-time partner DataStax, MongoDB competitor and leader in NoSQL and distributed databases. This is a strong acquisition, allowing IBM to bring DataStax’s Apache Cassandra-based technology to watsonx.
Years in the making, SK Hynix is finally closing its acquisition of Intel’s NAND flash, strengthening its presence in high-performance SSDs for enterprise and cloud computing. The deal, worth $8.84 billion, solidifies SK Hynix as a formidable competitor to Samsung in the NAND space
Transitions
Nokia has a new CEO, bringing Justin Hotard to the role. Hotard has a deep datacenter and AI background, coming to Nokia from Intel, where he led Intel’s AI efforts as EVP and GM for its Data Center & AI Business Group. He also spent nearly a decade at HPE where he focused on AI and HPC. This is a strong hire for Nokia, and big hint as to where their strategy is pointing.
There’s a new CMO at Cohesity, with the company naming Carol Carpenter to the position. Carpenter has previous stints as CMO at Unity and VMware, along with an executive role in marketing for Google cloud.
Cybersecurity company Barracuda Networks also has a new CMO. Peter Alexander joins the company after holding CMO stints at Fastly, Check Point, Harmonic, and Cisco.
What We’re Reading
So much for those embargos: the Wall Street Journal writes about how Chinese Buyers Are Ordering Nvidia’s Newest AI Chips, Defying U.S. Curbs.
While AI seems to be everywhere, not everyone is using it. The latest research from Pew Research shows that a more than 55% of employees “rarely or never use” AI chatbots, while an additional 29% haven’t heard of it at all, showing just how disconnected all us Silicon Valley types may be from what’s happening in the real world. Go deep on how workers look AI in Pew’s latest research on topic. Some of it will surprise you.
If their mothers couldn’t fix them… the Financial Times has an interesting read about how dating site Match enlists AI to nudge men into better behaviour on data apps:
“For ‘men especially’… When a user types an off-colour message, Match’s apps will generate an automated prompt asking them if they are sure they want to send it. ‘We think of it internally as too much, too soon,’ Roth said. A fifth of people who receive these prompts reconsider their messages, according to Match.”
Finally, Raconteur tries to answer the same questions your manager is asking, Sex, naps and Daytime TV: What do People Really get up to when Working from Home? It finds that “three in 10 admitted to switching off from work completely and taking a nap, while 15% have been intimate with a partner while working remotely.”
Not that any of this is bad (and, really, who isn’t more productive after a 2pm nap?), says a university researcher quoted in the piece:
“’I don’t see this as a productivity issue – I see it as people managing their life around their work,’ she says. ‘There is often a built-in assumption that employees who work from home will somehow take advantage or skive. However, there is almost no evidence that this is true. In fact, many people end up working longer days.’”
Companies mentioned: Acer, Barracuda Networks, Broadcom, Cisco, Cohesity, DataStax, HashiCorp, IBM, Meta, MongoDB, New Relic, Nokia, NVIDIA, OpenAI, SK Hynix, VAST Data, Voyage AI
