Welcome to this edition of our Weekly Analyst Insights roundup, which features key insights that our analysts have developed based on the past week’s events.
Competition for semiconductors to drive AI continues to heat up, and our analysis last week provided a few apt examples of that. First, Matt Kimball delivered his in-depth analysis of Google’s new TPU 8 chips for AI inference and training. Then we published a new research paper — written by Matt with an assist from Anshel Sag — on how Arm’s history of silicon IP design sets it up for the AI era, especially in the form of its new AGI CPU for the datacenter. Meanwhile, I appeared on CNBC to discuss AMD’s earnings and the market’s bullish reaction to the company’s AI-driven datacenter momentum. That segment also featured a discussion of Arm’s greatly expanded TAM (total addressable market) in datacenter chips.
Google’s new TPU 8i. (Credit: Google)
While NVIDIA often gets the lion’s share of attention for its AI chips — understandably so — this market continues to evolve quickly, and I am looking forward to seeing how things sort out, especially as datacenter power consumption becomes more of a gating factor in the next couple of years.
This week, Jason is attending SAP Sapphire in Orlando. The spring season of client and vendor events is in full swing, and we’re on the road a lot over the coming months — including Zendesk Relate, Google I/O, Dell Technologies World, and Computex this month. If you’ll be attending any of the same events or see that we’ll be in your city, please reach out.
Last week, Moor Insights & Strategy analysts’ perspectives appeared across a broad set of business and technology outlets, including Benzinga, Data Center Knowledge, SDxCentral, TechTarget, CIO, Barron’s, Seeking Alpha, and Armada International. Media coverage focused on AMD’s strong earnings, Astera Labs’ new Scorpio fabric switches for AI scale-up networking, and IBM’s enterprise AI operating model strategy and Bob coding agent launch. Analysts also weighed in on Intel’s stock surge tied to potential chip deals, Kopin’s financial turnaround and optical interconnect innovations, MongoDB’s enhanced vector capabilities for AI applications, and SAP’s strategic acquisition of data lakehouse vendor Dremio.
Our MI&S team also published 13 deliverables — 1 Forbes Article, 1 Research Paper, 6 Research Notes, 2 Analyst Insights, and 3 Podcasts.
Check out this week’s Analyst Insights roundup for more from the MI&S team, including what’s top of mind for each analyst, our thoughts on vendor announcements, press quotes, and more.
Have a great week!
Patrick Moorhead
IBM Think 2026: The Real AI Divide Is About Teams, Not Technology
IBM CEO Arvind Krishna opened Think 2026 with a clear thesis: The gap between AI winners and laggards isn’t about budget or model quality — it’s about how deeply AI is embedded in how teams work. IBM calls this the “AI Operating Model,” and three software products anchored the message: Bob (IBM’s agentic IDE positioned as a team coordination layer, with reported 45% productivity gains across IBM’s 80,000-developer workforce), Concert (an AIOps platform built to handle the operational scale of a world with billions of AI agents), and the Pearson-backed Enterprise Advantage model. Microsoft and ServiceNow launched competing frameworks the same week, which validates the direction.
What was challenging for me (and, it seems, for some others, based on anecdotal browsing of social media coverage) was how the high-level CEO messaging did not percolate into the product announcements. You really needed to be tuned in to catch the connections between the strategic and product narrative in the breakout sessions and in some of the keynotes. That led some to say the product aspect of the event was lacking. I appreciate that viewpoint especially if you were casually watching over the livestreams. And, yes, it is true that in many ways IBM’s software announcements demonstrated that its products are not very well differentiated from a feature perspective. That was a lost opportunity for IBM. But the underlying design philosophy — AI that serves team workflows rather than individual ones — is a meaningful distinction that IBM can leverage very well competitively. Ultimately, IBM’s future success will come down to whether it can deliver that operating model with the governance depth and hybrid deployment flexibility its enterprise customers require.
Read the full article → IBM Think 2026: The Real AI Divide Is About Teams, Not Technology
MongoDB .local London: It’s Not Just the Data — It’s the Context Infrastructure
Every AI vendor tells you that data is the foundation for good AI, but the real differentiator is the infrastructure that turns data into accurate, low-latency, real-time context at the moment an agent needs it. MongoDB’s .local London event made that argument with three concrete announcements: Automated Voyage AI Embeddings in Vector Search (public preview) that keep agent context current without manual pipeline work; MongoDB 8.3, which the company says delivers up to 45% more reads and 35% more writes at production scale; and the LangGraph.js Long-Term Memory Store reaching GA so agents maintain persistent memory across sessions.
I do like how Mongo is taking advantage of this moment when the market is so focused on AI and agents to reposition and elevate its value proposition to developers and other IT stakeholders. That said, it must also balance the temptation to pick a fight with the big data platforms and really carve itself a dominant position in next-generation agentic apps. Going after the big players risks dilution by trying to support older data design patterns that are becoming more and more commoditized.
With respect to last week’s announcements, two questions remain. First? How much of the benefit requires centralizing everything inside Atlas? Enterprises with existing multi-platform investments will push hard on interoperability. And, second, cost visibility at scale needs more clarity, as platform fees compound on top of cloud infrastructure costs in ways that aren’t obvious in a proof-of-concept. MongoDB CEO CJ Desai’s thesis — that the hardest part of running agents in production is the data layer, not the model — is correct. Now MongoDB needs to prove it can win that layer in heterogeneous environments.
Read the full article → It’s Not Just the Data — It’s the Context Infrastructure
IBM Think 2026: The Real AI Divide Is About Teams, Not Technology
IBM CEO Arvind Krishna opened Think 2026 with a clear thesis: The gap between AI winners and laggards isn’t about budget or model quality — it’s about how deeply AI is embedded in how teams work. IBM calls this the “AI Operating Model,” and three software products anchored the message: Bob (IBM’s agentic IDE positioned as a team coordination layer, with reported 45% productivity gains across IBM’s 80,000-developer workforce), Concert (an AIOps platform built to handle the operational scale of a world with billions of AI agents), and the Pearson-backed Enterprise Advantage model. Microsoft and ServiceNow launched competing frameworks the same week, which validates the direction.
What was challenging for me (and, it seems, for some others, based on anecdotal browsing of social media coverage) was how the high-level CEO messaging did not percolate into the product announcements. You really needed to be tuned in to catch the connections between the strategic and product narrative in the breakout sessions and in some of the keynotes. That led some to say the product aspect of the event was lacking. I appreciate that viewpoint especially if you were casually watching over the livestreams. And, yes, it is true that in many ways IBM’s software announcements demonstrated that its products are not very well differentiated from a feature perspective. That was a lost opportunity for IBM. But the underlying design philosophy — AI that serves team workflows rather than individual ones — is a meaningful distinction that IBM can leverage very well competitively. Ultimately, IBM’s future success will come down to whether it can deliver that operating model with the governance depth and hybrid deployment flexibility its enterprise customers require.
Read the full article → IBM Think 2026: The Real AI Divide Is About Teams, Not Technology
MongoDB .local London: It’s Not Just the Data — It’s the Context Infrastructure
Every AI vendor tells you that data is the foundation for good AI, but the real differentiator is the infrastructure that turns data into accurate, low-latency, real-time context at the moment an agent needs it. MongoDB’s .local London event made that argument with three concrete announcements: Automated Voyage AI Embeddings in Vector Search (public preview) that keep agent context current without manual pipeline work; MongoDB 8.3, which the company says delivers up to 45% more reads and 35% more writes at production scale; and the LangGraph.js Long-Term Memory Store reaching GA so agents maintain persistent memory across sessions.
I do like how Mongo is taking advantage of this moment when the market is so focused on AI and agents to reposition and elevate its value proposition to developers and other IT stakeholders. That said, it must also balance the temptation to pick a fight with the big data platforms and really carve itself a dominant position in next-generation agentic apps. Going after the big players risks dilution by trying to support older data design patterns that are becoming more and more commoditized.
With respect to last week’s announcements, two questions remain. First? How much of the benefit requires centralizing everything inside Atlas? Enterprises with existing multi-platform investments will push hard on interoperability. And, second, cost visibility at scale needs more clarity, as platform fees compound on top of cloud infrastructure costs in ways that aren’t obvious in a proof-of-concept. MongoDB CEO CJ Desai’s thesis — that the hardest part of running agents in production is the data layer, not the model — is correct. Now MongoDB needs to prove it can win that layer in heterogeneous environments.
Read the full article → It’s Not Just the Data — It’s the Context Infrastructure
Almost every event last week had governance in it somewhere. Some weeks it sits in the keynote, some weeks it shows up in the supporting product, but agents are pulling it forward in everyone’s pitches.
IBM’s Sovereign Core hitting GA at Think is the one I keep thinking about. It runs on Red Hat OpenShift on customer-owned infrastructure, so sovereignty becomes something you can actually procure and operate. Bringing Mistral in at the platform layer instead of forcing Granite was smart. Regulated buyers in Europe were never going to take a single-model story, and IBM didn’t ask them to. For CIOs in banking, healthcare, or government, this changes what the sovereignty conversation should sound like in your shop. I walked through the Sovereign Core announcement on LinkedIn, including the multi-model platform call and what it means for regulated buyers.
ServiceNow’s Knowledge announcements landed on the same theme. Autonomous Data Governance for observability, quality, and privacy is on the H2 roadmap, sitting next to RaptorDB Pro and Workflow Data Fabric. The autonomous part is what matters. Agents are already running against enterprise data, and the governance underneath them is still mostly manual. Humans aren’t going to keep up at agent speed, so something has to give. Say what you will about “autonomous” being overused everywhere today, but I think ServiceNow is right to call it that. Whether it ships on time and matches the pitch is the only thing I’m watching.
The infrastructure side of the market keeps shifting toward data. Storage vendors are pitching AI data platforms, HCI vendors are positioning themselves as agentic platforms, and the lines between infra, data, and governance keep moving.
Dell spent last week priming for DTW next week, and the run-up has a clear shape. ObjectScale picked up Wasabi for hot/cold tiering across AI data pipelines. PowerStoreOS got AIOps for update orchestration. The AI Platform expanded with new AMD-based PowerEdge configurations. The cyber resilience pieces fill out the data protection story without forcing rip-and-replace. The thread running through these is that Dell wants to keep enterprise data gravity on-prem in the AI era. If your data already lives on Dell, the pitch is that you can run agents against it without shipping it to a hyperscaler and without assembling the storage, ops, and protection layers yourself. That reframes the on-prem AI argument, especially for regulated industries that planned to keep data home anyway.
Nutanix has been chasing VMware defectors since the Broadcom acquisition, and Nutanix CEO Rajiv Ramaswami said as much on record this week. What changed is the upsell. Nutanix integration with Palo Alto Networks adds a governance and security leg under the agentic AI platform pitch, and the recent $150 million equity investment from AMD funds a deeper hardware path. Together, these move Nutanix’s conversations with VMware defectors away from “This is a cheaper place to run your VMs” and toward “This is the AI platform you’d land on anyway.” That matters because agentic buyers want governance baked in before they commit, and Nutanix needed the Palo Alto leg before that pitch held weight. The migration to Nutanix used to be about economics. The decision it’s now asking buyers to make is about which platform their next five years of AI workloads will run on.
At ServiceNow’s Knowledge 26 event last week, the company focused on how its platform is being extended with more autonomous and agentic AI capabilities across IT, CRM, employee, and security workflows, with new AI specialists designed to execute end-to-end processes rather than just assist with discrete tasks. The discussions and customer sessions highlighted that early outcomes are emerging in areas such as service resolution times, case deflection, and incident handling, while also reinforcing that clean processes and data remain prerequisites for reliable AI at scale. There was also a clear emphasis on security, both as a product area and as a growth vector, with ServiceNow positioning its security offerings as part of a broader approach to governing and coordinating AI-driven operations.
Ecosystem announcements, including an extended multi-year agreement with Lenovo that combines device intelligence and lifecycle management with the ServiceNow platform, and an Accenture initiative aimed at helping clients scale agentic AI implementations, pointed to a strategy that leans on partners to operationalize these capabilities in complex global environments. I will publish a deeper dive into Knowledge 26 and the implications of these announcements for enterprise buyers and partners in an upcoming research note.
I spent the front of last week in Redwood Shores, California, at Oracle’s Analytics and AI Summit. It gave me a chance to read the AI Data Platform, OCI Enterprise AI, and Fusion Data Intelligence stories side by side — how they fit, how they connect, and how much ground Oracle covers. The short version: More than I expected, more cohesive than the product mappings suggest. Newsletter coming on LinkedIn with the longer take.
At Knowledge 2026, ServiceNow tied together an impressive data story. RaptorDB Pro is now queryable via SQL through Live Connect; Live Archive pulls cold data into the same query at cheaper cost; and Workflow Data Fabric is GA with the Data Catalog. Add the Cloudera Zero Copy Connector and the Qlik and Pyramid partnerships, and you have a cost-optimized, performance-optimized foundation. For ServiceNow customers, the practical change is querying their data in place instead of extracting it to a warehouse, and reaching into outside data through the partner stack. I wrote up the full data-foundation read on LinkedIn, carved out from Mel Brue’s business apps coverage. The sales motion to data teams remains the open question.
Tableau and Teradata worked the same problem from different angles. The Tableau Conference unveiled the Agentic Analytics Platform on Tableau Semantics as the trusted knowledge layer that agents reason against. The repositioning matters because BI is getting commoditized by natural-language interfaces, and the durable position for an analytics vendor is owning the layer where business meaning lives. For customers, Tableau Semantics gives every agent and app one definition of revenue, customer, churn, and risk, so answers don’t drift across teams. Teradata launched the Autonomous Knowledge Platform across cloud, on-prem (Factory on Dell + NVIDIA), and hybrid, the on-prem leg being what Snowflake and Databricks can’t match for data-residency-constrained buyers.
IBM had tons of announcements at Think, but for the data world a few stand out. watsonx.data picked up GPU acceleration that delivered what the company said was a 30x price-performance jump and 83% cost savings, which changes the compute economics for running AI on enterprise data at scale. The Confluent integration plugs real-time streaming into the data layer. Agents need fresh data, and overnight batch loads don’t get them there. CEO Arvind Krishna’s AI Operating Model framing pulled agents, data, automation, and hybrid into one prescription, which is how regulated buyers actually buy.
SAP added two more acquisitions to make it three in six weeks: Reltio for master data, Dremio for an Iceberg lakehouse, and Prior Labs for tabular foundation models. Tied together, this gives customers agents reasoning across SAP and non-SAP data with trusted entities and models built for the math that an ERP actually runs. Extract-and-send-to-your-warehouse goes away. I broke down the three-acquisition stack on LinkedIn, with the Iceberg-native foundation and the European sovereign AI angle as the bigger calls. The big Sapphire event next week is where it has to land. Stay tuned for a research note on this in the next day or two.
Valve finally released its first of three new products, the Steam Controller. This isn’t the company’s first try at a controller, as it had previously made a very different Steam Controller entirely in-house and then discontinued it. This new Steam Controller, I believe, is designed to accompany the upcoming Steam Frame headset and Steam Machine desktop, all running on Linux. I believe the other two products are being impacted by the memory crunch, and pricing and specs might be in flux; meanwhile, the controller isn’t as affected by memory pricing, so I think it makes sense for Valve to release it now. The price of $99 isn’t necessarily low considering the current gaming market, but it does also seem justified considering the quality of the hardware and compatibility with Valve’s Steam platform and next-generation SteamOS. For much more context on this, see my December 2025 analysis of the Steam hardware lineup.
Samsung and Qualcomm have worked together to create big improvements in uplink performance in fixed wireless access deployments. This is significant because the new 5G Power Class 1 (PC1) capabilities are designed to enable better FWA coverage using existing 5G networks. They should particularly improve upload speeds, which will be critical for AI applications, where we are seeing people upload more content to the cloud and need strong upload performance more than ever. Samsung and Qualcomm claim that this achievement increases coverage by 40% and improves uplink throughput by up to 10x at the cell edge compared to the PC1.5 standard. This won’t necessarily translate to 10x across the entire network, but we could see users who previously got just 5 Mbps at the cell edge see 50 Mbps uploads from the same spot.
Microsoft’s latest Work Trend Index reinforces what I am seeing across enterprises: The real story is not “AI adoption” but whether organizations are willing to redesign work so humans and AI can actually collaborate. The data points to a growing disconnect between leaders and employees. Leaders say they want productivity and are investing in AI tools, but employees are still buried in meetings, notifications, and manual work that AI could absorb if workflows were intentionally rethought.
At the same time, the report highlights emerging frontier organizations that are moving beyond pilots and features to treat AI as a core orchestration layer for work, not an add-on. Those companies are seeing less anxiety about job loss and more confidence in AI as a collaborator because they are pairing deployment with training, new role definitions, and clarity about where humans create differentiated value. To me, the takeaway is that the competitive gap will not be between companies that “have AI” and those that do not. It will be between those that re-architect processes, metrics, and skills around AI-enabled workflows and those that simply bolt AI onto existing, already broken ways of working. I will be publishing a deeper dive on these findings and what they mean for enterprise leaders in an upcoming brief.
Researchers from Cleveland Clinic, RIKEN, and IBM have successfully simulated protein complexes containing up to 12,635 atoms — the largest biological molecular simulation to date created on a quantum computer. This is one of several recent quantum milestones to signal that quantum computers are entering an era of greater and more meaningful accomplishments. In other words, quantum is moving from a theoretical promise to practical achievements.
This breakthrough is especially important because it was achieved through quantum-centric supercomputing. By using that technology, we get a glimpse of a future where quantum processors and classical supercomputers routinely work together, each handling modality-specific tasks.
In this case, classical computers reduced complex protein-ligand structures into manageable fragments. The quantum piece was handled by IBM’s 156-qubit Heron quantum processors. The prototype quantum-centric supercomputer, which was housed at both Cleveland Clinic in the United States and RIKEN in Japan, calculated the quantum-mechanical behavior of each piece. The final results were then reassembled on two of the world’s most powerful supercomputers, Fugaku at RIKEN and Miyabi-G elsewhere in Japan. This produced a complete molecular picture.
This project has scaled rapidly. Six months ago, it could only handle molecules that were 40x smaller than they are today. A new hybrid algorithm, EWF-TrimSQD, enables simulations of previously inaccessible molecule sizes while improving the accuracy of a key step in the workflow by up to 210x.
Despite the excitement about this research, significant challenges still remain. Scaling and improving the performance of logical qubits, stability, and error correction remain issues for this and other quantum products. Simulations still rely on classical computing for most workflows. Bridging that gap will require continued algorithmic innovation alongside improved quantum hardware.
In the future, there are expectations to simulate enzyme catalysts, drug mechanisms, and other molecular behaviors that can currently be studied only through physical experimentation. In the long run, quantum-centric supercomputing will likely transform pharmaceutical research. This line of quantum research has the potential to accelerate new drug discovery, making it faster, cheaper, and more precise than before.
There is a practical motivation behind this research. We need to fully understand how a drug candidate binds to a target protein, which is one of the most difficult and expensive challenges in modern medicine. Current computational methods struggle with large molecules. As molecules grow larger, researchers must rely on approximations, which can result in costly errors late in the development process. Drug development timelines can stretch over a decade and require enormous investments. Improved molecular simulation could meaningfully compress that timeline.
NVIDIA’s deal with Corning should help to resolve a shortage within the AI industry around fiber optics, mostly caused by the need for fiber optics in networking for AI datacenters. NVIDIA will invest up to $3.2 billion to help Corning expand its fiber optic manufacturing capabilities, which the partners say will allow Corning to expand into three new facilities in the United States and create 3,000 jobs while increasing fiber optic capacity by 10x. This should also allow NVIDIA to continue to lean into CPO (co-packaged optics) for its new networking chips and potentially lower the cost of deploying fiber by guaranteeing supply. One possible side effect is that it could also help alleviate some of the shortages seen by fiber operators looking to deploy fiber internet.
The Wall Street Journal has reported that Apple plans to sign a chip deal with Intel Foundry to manufacture chips for its upcoming SoCs. While Apple and Intel may never publicly announce this deal, it seems material enough for the WSJ to report it, and the news has helped push Intel’s stock to a stratospheric level — one never before seen in the company’s history, in terms of both share price and market cap. I feel like this partnership was inevitable, given that Intel is a domestic U.S. foundry with industry-leading process nodes and the ability to help Apple multi-source its chips, potentially reducing costs — something I’ve written about extensively.
Unless otherwise noted, our analysts will be attending the following events in person.
August events coming soon.
September events coming soon.
December events coming soon.
Founder, CEO and Chief Analyst | + posts
Patrick Moorhead is the founder, CEO, and chief analyst of Moor Insights & Strategy. His big-picture view of technology is grounded in more than 20 years as an executive leading strategy, product management, product marketing, and corporate marketing functions at NCR, AT&T, Compaq, and AMD. He has shared his expertise in areas from silicon to infrastructure to enterprise SaaS and everything in-between in thousands of national broadcast appearances (CNBC, Yahoo Finance), articles (Forbes, CIO), research-based analyses, and podcast episodes. Today, he has 100+ CXO-level advisory clients and is often ranked the #1 technology industry analyst by ARInsights.
Anshel Sag is Moor Insights & Strategy’s in-house millennial with over 18 years of experience in the IT industry. Anshel has had extensive experience working with consumers and enterprises while interfacing with both B2B and B2C relationships, gaining empathy and understanding of what users really want. Some of his earliest experience goes back as far as his childhood when he started PC gaming at the ripe of old age of 5, building his first PC at 11, and learning his first programming languages at 13.
Mike Leone is a principal analyst at Moor Insights & Strategy covering data platforms and analytics, data infrastructure and storage, and data governance and enterprise data strategy. He brings 15 years of analyst experience from his work at Enterprise Strategy Group, where he rose to practice director for data management, analytics, and AI. Mike's work is grounded in a strong technical and strategic foundation, including early roles in software and hardware engineering.
Mel Brue is vice president and principal analyst covering modern work and financial services. Mel has more than 25 years of real tech industry experience in marketing, business development, and communications across various disciplines, both in-house and at agencies, with companies ranging from start-ups to global brands. She has built a unique specialty working in technology and highly regulated spaces, such as mobile payments and finance, gaming, automotive, wine and spirits, and mobile content, ensuring initiatives address the needs of customers, employees, lobbyists and legislators, as well as shareholders.
Matt Kimball is a Moor Insights & Strategy senior datacenter analyst covering servers and storage. Matt’s 25 plus years of real-world experience in high tech spans from hardware to software as a product manager, product marketer, engineer and enterprise IT practitioner. This experience has led to a firm conviction that the success of an offering lies, of course, in a profitable, unique and targeted offering, but most importantly in the ability to position and communicate it effectively to the target audience.
Senior Analyst-in-Residence | + posts
Bill Curtis is the Moor Insights & Strategy Analyst in Residence for large-scale Internet of Things systems. Bill helps enterprises design distributed solutions that integrate the full end-to-end IoT stack from real-world devices to analytics.
Jason Andersen is vice president and principal analyst covering application development platforms, technologies, and services. Jason brings over 25 years of experience in product management, product marketing, corporate strategy, sales, and business development at Red Hat, IBM, and Stratus to his work for MI&S and its advisory clients. Working both in the field and in the headquarters of some of the most innovative technology companies, Jason has a wealth of experience in building great products and driving their adoption across a broad spectrum of industries and use cases.
Paul Smith-Goodson is the Moor Insights & Strategy Vice President and Principal Analyst for quantum computing and artificial intelligence. His early interest in quantum began while working on a joint AT&T and Bell Labs project and, during 360 overviews of Murray Hill advanced projects, Peter Shor provided an overview of his ground-breaking research in quantum error correction.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。