Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Nvidia released new Cosmos and GR00T open models for both learnings and reasoning and a new framework for robotic training workflow. The company also outlined a strong ecosystem for autonomous machines and the availability of the Blackwell Jetson T4000 module.

At CES 2026, robotics was the big theme with a lot of hype focused on humanoid robots. Constellation Research analyst Esteban Kolsky argued in his newsletter that the focus on humanoid robots is a mistake. "Humanoid robots are the worst possible path we can take. Despite Hollywood’s love of anthropomorphized animatronics, there are many deficiencies in human-shaped and look-alike robots," said Kolsky in The Board: Distillation Aftershots.

Nevertheless, the combination of robotics and AI will be powerful and Nvidia and challengers such as Qualcomm all want to be dominant platforms. Robotics company Boston Dynamics, owned by Hyundai, outlined a collaboration with Google on models.

• A look at how Mercedes Benz, Nvidia collaborated on autonomous vehicles

During his CES 2026 keynote, Nvidia CEO Jensen Huang couldn't resist bringing out a cast of robots. The not so subtle message: Nvidia's platform powers robots from Boston Dynamics, Caterpillar, Franka Robots, Humanoid, LG Electronics and NEURA Robotics. Nvidia is also integrating its Isaac models and libraries with Hugging Face and LeRobot.

"Nvidia’s full stack of Jetson robotics processors, CUDA, Omniverse and open physical AI models empowers our global ecosystem of partners to transform industries with AI-driven robotics," said Huang.

Nvidia launched the following open robotics models.

• Cosmos Transfer 2.5 and Cosmos Predict 2.5, two world models that can be customized.
• Cosmos Reason 2, a reasoning vision language model.
• Isaac GR00T N1.6, an open reasoning vision language action (VLA) model for humanoid robots.

Huang pointed out robotics use cases for healthcare, manufacturing and construction via a collaboration with Caterpillar.

Nvidia is clearly going for scale. The company said its new Jetson T4000 module brings Blackwell to robots and can bring costs to $1,999 at a 1,000 unit volume. Huang highlighted a partnership with Siemens for physical AI and robotics.

"As the global labor shortage worsens. We need automation powered by physical AI and robotics more than ever," said Huang.

Qualcomm's robotics plan

Qualcomm at CES rollout out its robotics architecture that combines hardware, software and AI. Qualcomm also outlined its next-gen robotics processor for industrial robots as well as humanoids with its Qualcomm Dragonwing IQ10 Series.

In a nutshell, Qualcomm wants to be the energy efficient brain in robots that are mobile and autonomous. Qualcomm is also building a robotics ecosystem with partnerships with Figure, which makes humanoid robots, and other players including Advantech, APLUX, AutoCore, Booster, Figure, Kuka Robotics, Robotec.ai, and VinMotion.

Key items include:

• Qualcomm's Dragonwing industrial processor roadmap is designed to power multiple robotic form factors.
• The architecture supports perception, motion planning and AI models that enabled human interaction and generalized manipulation.
• Qualcomm has a bevy of developer tools for covering multiple robotics use cases for edge computing, software, machine learnings operations and what it calls the AI data flywheel.

The robotics play is part of the broader IoT plan for Qualcomm, which has built its broader IoT platform via the acquisitions of Augentix, Arduino, Edge Impulse, Focus.AI and Foundries.io. At CES, Qualcomm outlined its Q‑7790 and Q‑8750 processors designed to power on-device AI across multiple devices. Here's the stack.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Nvidia outlined its Alpamayo open AI models and datasets to bring reasoning to autonomous vehicles. At CES 2026, Nvidia CEO Jensen Huang said Mercedes Benz with Alpamayo will hit the road in the first quarter.

Huang said Alpamayo and its collaboration with Mercedes Benz is its first full-stack effort for autonomous vehicles (AVs). The approach with Alpamayo revolves around reasoning-based vision language action (VLA) models that bring human thinking to autonomous vehicle designs.

"Our vision is that someday, every single car, every single truck, will be autonomous," said Huang.

It's a vision shared by Ola Kallenius, CEO of Mercedes-Benz Group AG, spoke on a panel about the company's AV efforts with Nvidia.

"I had a chance with the combined Mercedes and Nvidia team to drive it through San Francisco and down into Silicon Valley is point to point navigation," said Kallenius, who noted the AV was operated by "a very sophisticated level two system.

Kallenius noted that the system with Nvidia is more like level 2 plus. "It feels that the car is on rails you're just driving, and it does everything. I drove uninterrupted for more than an hour through pretty heavy traffic," he said.

The general idea behind Alpamayo is to create a model and system that can handle novel or rare driving scenarios. Huang said Nvidia will support the models and AV systems on a long-term basis. For Nvidia, the AV effort is a bridge to robotics.

Nvidia released:

• Alpamayo 1, a chain-of-thought VLA model for AV researchers. The model is on Hugging Face and has a 10-billion parameter architecture. The model uses video input to reason and show its logic.
• AlpaSim, an open source end-to-end simulation framework.
• Physical AI open datasets with more than 1,700 hours of driving data.

Nvidia said Lucid, JLR, Uber and Berkeley DeepDrive were showing interest in Alpamayo. Huang added that Nvidia's full autonomous driving stack includes Alpamayo, a policy and safety evaluator, classical AV tools and Halos Safety OS.

According to Huang, the effort with Mercedes Benz took several thousand people and at least five years of work. "This entire stack is vertically integrated. Of course, in the case of Mercedes Benz, we built the entire stack together. We're going to deploy the carbon, operate the stack, and maintain the stack for as long as we shall live," said Huang.

Kallenius said the Nvidia collaboration with Mercedes Benz revolves around safety.

"If you're moving an object that weighs 4,000 pounds, and it's moving at 50 miles an hour, sorry is not going to cut it. There needs to be a higher level of certainty and safety. You don't have to rush into the market. You don't have to be the first but you got to make sure that what you do is robust," said Kallenius.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Samsung and Munich Re's Hartford Steam Boiler (HSB) unit, which offers insurance for equipment breakdown, are planning to leverage IoT sensors, AI and smart appliances to spin up personalized home insurance. Welcome to the 2026 monetization of AI use cases.

The model was outlined as part of Samsung's First Look presentation at CES 2026, which included a plan that revolves around connecting and embedding on-device and cloud AI throughout its appliance lineup with a unified OneUI experience. More thoughts on that later including my working theory that consumers will pay premiums to have dumb appliances that just do what they're supposed to and are reliable.

But the HSB-Samsung partnership is worth noting because it's a financial incentive that may prod consumers to bring more AI into their everyday experiences. Now you may not need or want Bixby as an AI experience or Google Gemini powered AI vision to examine the food in your fridge or AI recipe recommendations, but with some financial incentives you might. The catch: HSB insurance appears to mean that everything in your house needs to be connected via Samsung's SmartThings.

Based on Samsung's presentation, even your steam wrinkle iron needs a Qualcomm Dragonwing processor.

However, the HSB partnership with Samsung is going to expand beyond a pilot in the US to more states. HSB President and CEO Greg M. Barats said smart appliances connected to SmartThings can lower risk and premiums.

"Using information from these connected appliances, we assess the characteristics of your smart home and enable a truly personalized offer, knowing that you have appliances smart enough to alert you a small leak before it floods your kitchen, we can help you avoid costly, time consuming claims."

In a nutshell, HSB and Samsung want to apply the industrial AI preventative maintenance use case to your home.

HSB in 2025 launched its Meshify Defender Slim sensor, powered by Amazon Sidewalk, to monitor buildings and homes for water leaks and frozen pipes. HSB and Flume also offer water leak protection.

This pilot with HSB and Samsung is worth watching since it highlights how AI may be monetized going forward. Use of AI and models in insurance is also notable. 

• Neptune Insurance highlights AI exponential, data, build vs. buy trends
• BT150 interview: Doug Benalan, CIO Cure Insurance
• Cognizant: AI savvy consumers will drive spending, demand new experiences

Thoughts on Samsung's AI strategy

Samsung's First Look included a 130-inch TV, an AI-driven Family Hub refrigerator, a Bespoke AI Laundry Combo that eliminates the need to transfer loads of laundry and a Bespoke AI Jet Bot Steamer.

TM Roh, CEO and Head of Samsung’s Device eXperience (DX) Division, said:

"Our mission is clear to be your companion to AI living, and our strategy is simple. We will harness the full scale of Samsung to create technologies and experiences that surely matter to people. We will embed AI across every category and every product and every service to deliver one seamless, unified AI experience. Our device portfolio puts us in every moment of daily life, giving us unique consumer understanding."

Here's the opportunity and the problem:

• Samsung ships more than 500 million devices each year and has a treasure trove of consumer data. But my confidence that Samsung can provide one unified experience isn't high. Samsung can't resist layering its Bixby assistant on top of Gemini. App conflicts are everywhere on a Samsung smart phone to the point where I ditched Samsung for Google's Pixel and Motorola's Razr. I don't need that UI disaster on my fridge.
• It's not clear anyone is asking for AI features on their appliances. I just want my washer and dryer to work without some Samsung sensor freaking out. Confession: My Samsung dryer is a royal pain and works once in a while. If the Samsung appliance didn't come with the house it wouldn't make the cut.
• AI everywhere sounds interesting on stage, but annoying in real life.
• Ultimately, people will pay up for dumb appliances. Some appliance maker should just market with a line like: "Dumb appliances that just do what they're supposed to do reliably. Every time. For years."
• Samsung has a good security ground game with Samsung Knox and on-device AI, but it's unclear whether consumers are going to be into sharing data with insurers. The data sharing will depend solely on the size of the discount. Details on the HSB discounts were sparse in Samsung's CES 2026 keynote.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Enterprises will get all-you-can eat agentic AI pricing, data tools are going to be a headache, AI agents will look more like a feature than a revolution and physical AI will matter. Those are some of the trends to watch in 2026.

Here’s a look at the trends and predictions in enterprise technology for 2026 grouped by confidence levels.

High confidence

Agentic enterprise license agreements will become the norm as CxOs push back. In 2025, enterprise software vendors, which were worried about losing seats, introduced consumption models. The idea was that a hybrid approach would be the best of both worlds for CIOs. However, consumption models were unpredictable and CIOs, not to mention CFOs wanted predictability. Enter the agentic AI enterprise license.

Salesforce's Miquel Milano, president chief revenue officer, laid out the rationale behind Salesforce's AELA, Agentic Enterprise License Agreement. "AELA is for customers that have already experimented. They're ready to scale. They want to go all in so we agree on a flat fee, and then it's a shared risk," said Milano.

ALEA is all you can eat with Agentforce or whatever other cloud is thrown in. Of course, once you're all in Salesforce can better monetize the platform next contract. In 2026, it's highly likely you'll see similar arrangements from SaaS providers.

Here's the catch: SaaS providers may ink ALEAs at a loss as they play for the renewal (when you're completely locked in). Milano is looking at lifetime value of a customer. "If the customers are smart, they can rob the bank. They can really make a great deal out of that. We take the risk because we want our customers to be successful. I would love to have a customer where I price an AELA at $5 million incremental, and the customer has deployed so much that the deal is not profitable for me. If the deal isn't profitable for me, it means that the customer is the happiest customer in the world. And then I have another 20 years to monetize that customer," said Milano.

Enterprise data tolls and API economics are going to be a headache. Celonis is suing SAP over data access. In October, U.S. District Judge Vince Chhabria in San Francisco ruled the SAP must face the Celonis lawsuit. Toward the end of 2025, The Information reported that Salesforce was raising prices on apps that tap into its data. CIO.com noted that these connector fees are likely to trickle down to IT budgets.

• Celonis, SAP reach data access cease fire amid litigation
• Enterprise AI: It's all about the proprietary data

As agentic AI is deployed and agents connect, there will be multiple skirmishes in 2026 over these data tolls. There will need to be some standard toll to cover compute costs, but enterprises need to keep in mind that they own the data. In some cases, your vendor may feel otherwise.

Connection fees are going to be the new cloud egress to move data. I'd argue that data fees are going to be the biggest risk to scaling AI agents.

Agentic AI is a feature and the real deal is decision velocity. This theme is Michael Ni's department, but decision velocity--how quickly smaller decision trees and processes can be automated at scale--will be a core theme. When Ni, in a video chat, said agentic AI was merely a feature I had two reactions. First reaction: How many pixels have been spent on a feature?!? Second reaction: Mike has a point. The game is decision velocity and you're task in 2026 will be to put the pieces in place to get there.

"What we're seeing is the first decisions are smaller decisions, automations on the backside of that leading to human engagement. Then we start to collapse these decision trees to get to some of the 5x, 10x improvements. We're seeing leaders actually able to achieve with decision automation," said Ni, who noted that 2025 was a year where a lot of the decision velocity building blocks were put in place.

In-house forward-deployed become a necessity. Software vendors—all cribbing a play from Palantir—talked up forward deployed engineers almost every time executives talked agentic AI deployments. In 2026, enterprises are going to realize they need their own forward-deployed engineers to work through data, process, architecture and AI automation. These engineers will know the business and industry better than the ones you’ll borrow from software vendors and services firms.

Medium confidence

The AI market will bifurcate as the bubble pops either in 2026 or 2027. Concerns about AI infrastructure capital expenses and debt will be what really scales in 2026. But don't get distracted with your AI plans. Chances are 2026 will highlight how the easy money in training LLMs, raising capital at a ridiculous clip and fabulous remaining performance obligations is over. That AI market, dominated AI's circular economy and OpenAI, will be completely different than the enterprise AI version. That productivity boom is just starting as AI and process automation converge.

• See: Watercooler debate: Are we in an AI bubble?

Build will beat buy. The build vs. buy debate continues as AI agents make it easier to create applications you used to buy. In addition, build looks like a great option since customers are beginning to push back on SaaS deal inflation. SaaS costs go up annually. It's almost as bad as healthcare.

In 2026, there will be an inflection point where enterprises become convinced that applications custom to their use cases are the way to go. Sunk costs in enterprise systems will be abstracted away using agentic AI as a user interface.

• The road to AI Exponential will be bumpy
• Rimini Street’s second act will include heavy dose of agentic AI, UX
• CCE 2025: AI agents: Dreams, reality and what's next

AI benefits broaden out to more levels of the enterprise stack. Software vendors are going to show revenue and productivity gains from AI. On Wall Street, enterprise software vendors finally join the AI rally. Nvidia and AI infrastructure plays will be flattish in 2026.

Physical AI gets its moment. Manufacturing and industrial sectors start to leverage physical AI to deliver real value. In addition, physical AI breakthroughs will begin to rival the early days of LLMs. This physical AI focus drives optimism about robotics as well as edge AI applications.

Not absurd, but unlikely for 2026

• Nvidia shares close the year flat to down year, but the fall isn't enough to result in real selling. Sales growth begins to slow as hyperscalers increasingly rely on their custom silicon. There's a concentrated effort in the industry to break Nvidia's hardware and software moat around AI.
• AI infrastructure is overbuilt due to hardware and software advances that don't require as much compute and energy. Wall Street rewards companies like Apple that didn't go crazy building AI infrastructure. AI backlash builds as multiple NIMBY grassroots efforts thwart plans to build data centers in small towns and rural areas. Note I said something similar to this in 2025 to no avail.
• OpenAI realizes it can't continually raise money forever goes on with an austerity push to show it can generate cash flow and become profitable.
• Quantum use cases go mainstream in the enterprise as quantum supremacy comes early. It also becomes clear that superconducting quantum computing is the clear technology winner and that reality causes a mad scramble for companies focused on trapped ions, neutral atoms, annealing and other techniques.
• Meta revamps its AI operations again once it's clear that its new management team and focus doesn't yield results. Meta's AI operations resemble the New York Mets, a massive payroll that doesn't deliver wins.
• High memory costs create a buyers strike for PCs, servers and smartphones.
• 2026 becomes one of the biggest years ever. for IPOs as Databricks, Anthropic, OpenAI, SpaceX and Stripe all go public. Two of those five headliners trade under their IPOs prices after 3 months.
• TikTok usage plunges under new ownership as the US algorithm is tweaked.

Scorecard from 2025

Here's a look at my predictions for 2025 that worked out and other areas where my crystal ball was cracked. In 2025, I included probability of a prediction playing out.

On target:

• 2025 was more volatile than usual.
• AI productivity gains broadened in enterprises.
• OpenAI and Microsoft became frenemies without a doubt.
• Enterprises will try new SaaS revenue models and scream. Consumption sounds great...until you get the bill.
• ERP comes under fire. ERP didn't go anywhere but is being abstracted.

Off base:

• That alleged run on vendor consolidation never played out.
• Agentic AI will usher in autonomous processes. We're clearly still in the early stages. The buy side is still wary of lock-in.
• The AI data center buildout will stall. Actually, it just accelerated in a few days after I made that prediction.
• Nvidia growth slows. Nvidia still faces the law of large numbers and more competition, but 62% revenue growth is pretty sweet.
• Edge computing becomes more critical for AI workloads. It'll happen but sure didn't in 2025.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

DeepSeek published a technical paper co-authored by co-founder Liang Wenfeng that argues for a new architecture to train foundational models.

The paper, which details Manifold-Constrained Hyper-Connections (mHC), argues that a new architecture is needed to better scale deep learning without signal divergence. Granted, the mHC paper is wonky, but the key takeaway is that you can have large-scale training that's more efficient with better quality.

"Empirical results confirm that mHC effectively restores the identity mapping property, enabling stable large-scale training with superior scalability compared to conventional HC. Crucially, through efficient infrastructure-level optimizations, mHC delivers these improvements with negligible computational overhead."

Given that DeepSeek's paper landed the day after New Year's Day, it's unlikely to garner a ton of attention. However, DeepSeek's paper is just another development leading to the following: The muscle head era (error) of AI is coming to an end.

The AI market to date has been driven by muscle head logic. The answer for everything in the last two years has been more muscle. In AI's case that approach means more GPUs, more farm land going away, more energy usage, more water and by all means more data centers, debt and capital expenditures.

Almost a year ago, DeepSeek's LLM rattled the AI market because it appeared to be more efficient and trained without the latest and greatest from Nvidia. After an early freakout over DeepSeek, the US-driven part of the AI market went right back to the Stargate happy approach. Toward the end of 2025, investors were starting to question the capital expense and debt load behind the AI buildout.

In 2026, we're likely to see more engineering excellence and less muscle head logic. Here's some evidence to why that theory may play out.

• Price and performance matter. LLM giants were busy diversifying their chip bases. Google's TPUs and Amazon Web Services' Trainium are going to be threats to Nvidia as inference becomes the main event.
• Capital expenditures and debt loads are being questioned. Just ask Oracle and Meta.
• Nvidia's not acquisition of Groq. Nvidia licensed Groq and hired damn near the whole company away because Groq's approach to memory and inference is more cost effective.
• NIMBY is brewing. Yes, tech titans would pave over every inch of land for an AI factory, but you need power. Communities are going to become wary of paying higher electric bills for AI. Sen. Bernie Sanders of Vermont and Florida Gov. Ron DeSantis agree on little, but both agree AI's data center boom is a raw deal.
• China is pushing new architectures that are easy on compute. The US has deprived China of Nvidia's best chips (until recently). As a result, China has eyed more efficient training and embraced open source. DeepSeek has the incentive to push new AI architectures. The US is more muscle over elegance.

DeepSeek's real legacy: Shifting the AI conversation to returns, value, edge | China vs. US AI war: Fact, fiction or missing the point?

These developments mean you can't just see DeepSeek's paper in a vacuum. It's just one more piece of evidence that AI is going to become way more efficient and that potentially means less compute.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Meta acquired Manus in a deal that aims to give the company a rare win in artificial intelligence after a year of revamping. Manus is critical to fleshing out Meta CEO Mark Zuckerberg's dream of a universal AI assistant to get work done across its products.

Whether Meta allows Manus to run or bogs it down remains to be seen, but there are some fun facts worth noting.

• Manus is on a $125 million annual revenue run rate after eight months.
• The company is growing 20% month-over-month since it releases Manus 1.5.
• Manus' portfolio of agents is focused on getting work done including making presentations, design and a laundry list of other things SMBs need to do. Manus has multiple connectors to various systems.
• The company released its General AI Agent earlier in 2025 and the goal is to enable users to delegate tasks.
• In many respects, Manus was able to move faster than the giants with AI agents that actually deliver results for various use cases.
• Manus plans range from $20 a month to $200 a month with Team plans available.

Yes, Manus flew under the radar relative to OpenAI and Anthropic, but can accelerate Meta's plans. What are Meta's AI plans other than handing out massive pay packages, spending big on capital expenditures and falling behind on LLMs? Here's what Zuckerberg said in October about the company's AI research spending and agentic AI.

"The research is going to enable new technological capabilities to exist. And then those capabilities can get built into all kinds of different products. So the ability to reason more intelligently is, for example, very important across a large number of things. It would be useful for an assistant. It will also be useful in business AI. It will also be useful in the AI agent that we're building to help advertisers figure out what their campaigns are going to be....

There are lots of different capabilities to build. I'm not sure that any one company is going to be the best at all of them. I doubt that's going to be the case. But a lot of what we're trying to do is not like -- not kind of do some things that others have done. We're really trying to build novel capabilities."

In July, Zuckerberg said:

"I believe every business will soon have a business AI, just like they have an e-mail address, social media account and website. Our focus is now deepening the experience and making Meta AI the leading personal AI. As we continue improving our models, we see engagement grow."

• Meta Superintelligence Labs: A look of the challenges ahead

Manus should be able to get Meta closer to its AI agent vision. Meta said in a statement that it will "we will continue to operate and sell the Manus service, as well as integrate it into our products."

The paths forward for Manus could go like this:

• Meta will run Manus independently and it'll become a business AI brand that scales to a multi-billion revenue unit. Meta has built out its Instagram and WhatsApp businesses as part of a portfolio so there's some precedent.
• Manus agents will be built into a series of Meta products and absorbed. The technology remains, but Manus as a standalone business fades.
• Meta both operates Manus as a business and leverages it internally to maximize returns. Manus and Meta iterate on products and services.

Xiao Hong, CEO of Manus, said "joining Meta allows us to build on a stronger, more sustainable foundation without changing how Manus works or how decisions are made."

For now, Meta's Manus acquisition puts it back into the AI conversation as something more than a big spending also-ran.

Constellation Research's take

Holger Mueller, an analyst at Constellation Research, said the Manus deal is more of a hedge against its core advertising business. It's also an agentic AI play, but don't forget the ad angle.

Mueller said:

"In a world where it is not clear how AI is going to change how ads are going to be presented, consumed and paid for, Mark Zuckerberg is doubling down on AI and human supervision. The Manus agents pose interesting automation potential for knowledge workers, and an advertisement funded version of Manus, delivering AI automation first and then the ad at the moment of supervision. Brownie points for Meta if they are context and action aware."

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Quantum computing hogged the headlines in 2025 and it was ok to say it was the year of quantum--or maybe qubits--after just a few months. The quantum computing developments were flying, but it's worth noting that we're years away from big commercial adoption.

Nevertheless, CxOs need to get ready. After all, the boardroom is getting tired of AI. The AI trade lost steam. Boardrooms are going to start asking about your quantum computing plans in 3, 2, 1.

Why was 2025 the year of quantum? For starters, there was a new development almost weekly. Pure play quantum stocks were hot. Hyperscale cloud players were deadly serious about quantum, with AWS, Microsoft Azure and Google Cloud all running credible efforts. Quantinuum threaded the needle between AI and quantum computing. IBM scaled aggressively. And real use cases emerged as companies like IonQ cited projects with DARPA, AstraZeneca, and others.

Here's what's happened in 2025.

January

• Nvidia CEO Jensen Huang said quantum computing is 15- to 30-years away from being really useful. Those comments landed at CES 2025 and came up because quantum computing stocks had a torrid run at the end of 2024 as quantum became more practical, focused on hybrid HPC architectures and Google Cloud touted a new quantum chip called Willow.
• Quantum vendors have tried to prove Huang wrong since he made those comments. Days after Huang’s comments the pure plays such as IonQ and D-Wave all cited commercial projects.
• Microsoft launched its Quantum Ready Initiative, which helps enterprises sort out use cases for quantum computing and get ready to deploy.
• IonQ, University of Maryland and State of Maryland announced a $1 billion investment to create the "Capital of Quantum." Under the initiative, IonQ will be an anchor partner as it continues to pursue federal government contracts. IonQ at the University of Maryland's Discovery District, which is near Washington, DC and Annapolis, MD.
• D-Wave said it will expand distribution of its quantum computing offerings via a partnership with Carahsoft, which is a premier distributor of IT to the government. Under the deal, Carahsoft will aggregate D-Wave's offerings as a reseller. The partnership will give D-Wave throughput under Carahsoft's contract vehicles.

February

• Quantinuum launched its Generative Quantum AI framework that aims to combine AI, quantum computing and supercomputers to address problems classical computing can't solve.
• Microsoft launched Majorana 1, a quantum computing chip with a Topological Core architecture. Topological quantum computing uses a concept similar to semiconductors using "anyons," which can arrange qubits into patterns. A topological superconductor is a material that can create a new state of matter. It's harnessed to create a more stable qubit that can be digitally controlled.
• IonQ is doubling down on quantum networking with an acquisition, diversifying globally and has a new CEO along with fourth quarter results and plans to raise more capital. The purchase of ID Quantique gives IonQ more quantum networking patents and diversifies the company's global footprint with expansion in South Korea. In November, IonQ bought Quibitekk. IonQ named Niccolo de Masi CEO, with Peter Chapman becoming Executive Chair. Chapman will focus on strategy and enterprise adoption.
• Amazon Web Services launched Ocelot, a first-generation quantum computing chip that was developed with the California Institute of Technology.
• And for good measure, Constellation Research updated its quantum computing shortlists. Constellation ShortList™ Quantum Computing Platforms | Quantum Computing Software Platforms | Quantum Full Stack Players
• 

March

• D-Wave said its quantum computer outperformed a classical supercomputer in solving magnetic materials simulation problems. The company also reported its fourth quarter results. D-Wave followed up with a quantum blockchain architecture.
• Nvidia CEO Jensen Huang at Nvidia GTC 2025 hosted the company’s first quantum day. A panel that featured most of the quantum computing CEOs amounted to a mea culpa tour following Huang’s comments in January.
• Quantinuum quantum computers have created true verifiable randomness in a project that could be valuable to cybersecurity. In a paper in Nature, Quantinuum along with JPMorganChase, Oak Ridge National Laboratory, Argonne National Laboratory and the University of Texas have generated true randomness critical to cryptography and cybersecurity. Quantinuum said the latest advance was built on research from Shih-Han Hung and Scott Aaronson of the University of Texas at Austin.
• IonQ and Ansys said they also outperformed classical computing when designing medical devices. IonQ also expanded its quantum networking portfolio and raised more than $372 million via an equity offering.
• Classiq said it worked with Sumitomo Corp. and Mizuho-DL Financial Technology to successfully compress quantum circuits in implementing quantum algorithms for Monte Carlo simulations.
• Rigetti Computing and Quanta Computer said they will spend a combined $500 million to accelerate development and commercialization of superconducting quantum computing. The company also reported fourth quarter revenue of $2.27 million and 2024 revenue of $10.79 million. The net loss for 2024 was $200.99 million.

April

• IBM said it will spend $30 billion in R&D in the US as part of a broader $150 billion spend. That R&D in part will be devoted to quantum computing. IBM CEO Arvind Krishna said on the company’s first quarter earnings call: "In Quantum, we are proud to partner with the Basque Government to deploy Europe's first IBM Quantum System 2 in Spain, a milestone in global Quantum leadership." IBM also published a paper on how quantum addresses problems in combinatorial optimization.
• IonQ moved to expand its reach. It established a quantum computing and networking hub in Chattanooga, Tennessee in a $22 million deal. The company also outlined an agreement with Toyota Tsusho and AIST to expand quantum computer reach in Japan. IonQ also signed a memorandum of understanding with Intellian to extend into South Korea. IonQ also said its Forte Enterprise system is available through Amazon Braket.
• D-Wave and Davidson Technologies, a defense technology company, assembled D-Wave's Advantage2 annealing quantum system at Davidson's Huntsville, Alabama headquarters.
• Classiq, a quantum software development company, and Wolfram Research joined CERN’s Open Quantum Institute (OQI). The two companies will target quantum-based technologies to optimize electrical networks.
• Researchers at Tokyo University of Science, Japan unveiled DSAPS, a chip system that overcomes capacity and precision limits with dual scalable annealing processors.
• U.S. Defense Advanced Research Projects Agency’s (DARPA) Quantum Benchmarking Initiative (QBI) selected host of companies to evaluate a variety of technologies for creating quantum bits. The companies include: Alice & Bob, Atlantic Quantum, Atom Computing, Diraq, Hewlett Packard Enterprise, IBM, IonQ, Nord Quantique, Oxford Ionics, Photonic Inc., Quantinuum, Quantum Motion, QuEra Computing, Rigetti Computing, Silicon Quantum Computing Pty. Ltd. and Xanadu.

May

• IonQ went shopping again with the acquisitions of Lightsynq Technologies, a startup focused on photonic interconnects and quantum memory, and Capella Space, which specializes in quantum space networks. See: IonQ’s plan: Quantum networks extending into space
• IBM launched a Flex Plan for access to its quantum computing hardware in a move that aims to expand access for organizations, enterprises and researchers who want access without monthly limits. The IBM Quantum Flex Plan provides access on day one with an entry point of $30,000+ to gain access to Big Blue's entire fleet of quantum systems. Flex plan users will also get the advanced software features, support and early access to new releases.
• Classiq, a quantum computing software company, said it raised $110 million in Series C funding. The company is looking to build the software stack for quantum computing. The funding, led by Entrée Capital and a bevy of other investors, will be used to build Classiq's go-to-market, customer success and R&D teams globally.
• D-Wave Quantum said its Advantage2 annealing quantum computer is now commercially available and is likely to contribute to revenue growth. Advantage2 is available via D-Wave's Leap quantum cloud service as well as on-premises deployments. Dr. Alan Baratz, CEO of D-Wave, said the system is a milestone in the company's development and able to "solve hard problems outside the reach of one of the world’s largest exascale GPU-based classical supercomputers."
• Cisco is entering the quantum computing networking ring with a lab and prototype processor. The move is notable since Cisco is a networking giant in the enterprise. In addition, quantum networking has been seen as a key piece of infrastructure.

June

• IonQ said it will acquire UK's Oxford Ionics in a deal valued at $1.075 billion in mostly stock and $10 million in cash. The deal is designed to accelerate IonQ's quantum computing roadmap and establish a global hub for research and development.
• Quantum Computing Inc. raised $200 million with an At-the-Market equity offering.
• IBM and Riken, a national research laboratory in Japan, unveiled the first IBM Quantum System Two ever to be deployed outside of US. The quantum computer will be co-located with Riken's supercomputer Fugaku.
• Rigetti completed a $350 million At-the-Market equity offering.
• IBM updated its quantum computing roadmap heading into IBM Quantum Starling, a large-scale fault-tolerant quantum system in 2029. Big Blue said IBM Quantum Starling will be delivered by 2029 and installed at the IBM Quantum Data Center in Poughkeepsie, New York. That system is expected to perform 20,000 times ore operations than today's quantum computers.

July

• D-Wave raised $400 million with an At-the-Market equity offering. The move left D-Wave with about $815 million in cash. D-Wave also outlined a strategic development effort to advance cryogenic packaging.

August

• D-Wave released a collection of developer tools for quantum AI and machine learning.
• Rigetti Computing said its 36-qubit multi-chip quantum computer is generally available and outlined a plan to build a 100-qubit system with 99.5% fidelity by the end of 2025, but quantum advantage is likely 4 years away. Second quarter results fell short of estimates.
• IonQ, which has been on an acquisition spree, reported second quarter results with a net loss of $177.5 million on revenue of $20.7 million. IonQ had $1.6 billion of pro forma cash and equivalents as of July 9. IonQ also filled out its leadership team with David Chung as VP, Corporate Development; Shad Reed as VP, Engineering Public Sector; Petrina Zaraszczak as VP, Business Operations & Integration; and Sterling Zumbrunn as VP, Product Management Networking.
• IBM, AMD eye hybrid quantum computing, supercomputer architectures.
• Microsoft rolls out quantum safe plan, timelines.
• Quantinuum fleshes out quantum software stack with open source Guppy, Selene.

September

• Quantum networking coming into focus with IonQ and Cisco playing roles.
• Microsoft said it will build quantum center in Maryland.
• Rigetti said it landed two orders for quantum computing systems for $5.7 million. Rigetti also announced a $5.8 million AFRL contract with QphoX.
• Quantum Computing closed a private placement of common stock to raise $500 million.
• IonQ announced purchase of Vector Atomic, completed the Oxford Ionics acquisition and inked a US Department of Energy memorandum of understanding to deploy quantum technology in space.
• IonQ launched a federal unit to focus on US government use cases. The company also named Inder M. Singh as Chief Financial Officer and Chief Operating Officer.
• SPACs are back as Horizon Quantum said it would go public via a merger with dMY Squared Technology Group.

October

• Quantum computing companies, often with minimal or no revenue, raise funds at a brisk pace to fortify balance sheets. Raise money when you don't need to is a good motto for CFOs to have. Even after a recent pullback Rigetti shares are up 200% on the year and IonQ is up 50%. D-Wave is up 356%. 
• Swiss Quantum Technology inked a €10 million to deploy D-Wave's Advantage2 system.
• The Basque Government and IBM launched Europe's first IBM Quantum System Two in Donostia-San Sebastián.
• IonQ priced a $2 billion equity offering to fortify its balance sheet following a run-up. IonQ also completed its acquisition of Vector Atomic.
• Arqit Quantum said its second quarter revenue wil be $460,000 to $470,000 with fiscal 2025 revenue of $525,000 to $535,000. The company had $36.9 million in cash.
• Quantum Computing announced a private placement of common stock to raise $750 million.
• Google said its Willow quantum computing chip has achieved quantum advantage. Google's breakthrough hit pure play quantum computing companies.

November

• Quantinuum launches Helios quantum computer, touts fidelity, enterprise customers
• IBM outlined its roadmap and launched its Nighthawk quantum processor. Big Blue also partnered with Cisco on quantum networking.
• IonQ, Rigetti, and D-Wave reported earnings as the pure plays continue to battle the giants.
• IonQ named Scott Millard Chief Business Officer and acquired Skyloom Global.

December

• IonQ and KiSTI finalized plans to bring a 100-qubit quantum system to South Korea.
• D-Wave said it will outline its commercial quantum plans at CES 2026.
• Quantum Computing Inc. named Dr. Yuping Huang CEO and acquired Luminar’s photonics business for $110 million.
• D-Wave launched a US government unit.

What now?

One thing worth noting here is that 2025 became the year of quantum readiness and development rather than deployment. That said, CxOs need to start thinking about quantum. Constellation Research analyst Holger Mueller broke down how you should be thinking about the year of quantum.

• If you need to protect IP against state actors, you should deploy quantum key encryption today.
• If you have smaller planning and simulation issues in your business, get your pilot going.
• If you are on process manufacturing, chemical, pharma etc. - you must have quantum pilots. You likely have them on annealing and laser gate already but need to keep an eye on all super conducting players.
• Everybody else can sit back and get the popcorn, but keep an eye on medium-scale planning and simulation use cases. All enterprises have those. On the tech side, watch stabilization, componentization and then scale loop. IBM is the most advanced and learn from their process.

Larry Dignan

Editor in Chief of Constellation Insights Constellation Research

About Larry Dignan: Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning. He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.&nbsp; <br>Constellation Insights does the following: Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and… Read more

Groq announced it has entered a non-exclusive licensing agreement with Nvidia for Groq's inference technology. Groq CEO Jonathan Ross and team will join Nvidia, but Groq will operate independently with GroqCloud.

News of a licensing deal landed as CNBC reported that Nvidia was buying Groq.

CNBC reported that Nvidia is acquiring Groq for $20 billion. CNBC's David Faber cited Groq investor Alex Davis, CEO of Disruptive, has his primary source. Groq, founded in 2016, had a valuation pushing $7 billion.

Groq’s secret sauce is a processor it calls the LPU (language processing unit). ‘The LPU integrates hundreds of MB of SRAM as primary weight storage (not cache), cutting latency and feeding compute units at full speed. This enables efficient tensor parallelism across chips, a practical advantage for fast, scalable inference,” said Groq. Groq's LPU is designed specifically for efficient AI inference.

Following the CNBC report, Groq announced it has entered a non-exclusive licensing agreement with Nvidia for Groq's inference technology. Groq statement in full:

“Today, Groq announced that it has entered into a non-exclusive licensing agreement with Nvidia for Groq’s inference technology. The agreement reflects a shared focus on expanding access to high-performance, low cost inference.

As part of this agreement, Jonathan Ross, Groq’s Founder, Sunny Madra, Groq’s President, and other members of the Groq team will join Nvidia to help advance and scale the licensed technology.

Groq will continue to operate as an independent company with Simon Edwards stepping into the role of Chief Executive Officer.

GroqCloud will continue to operate without interruption."

The Groq-Nvidia deal rhymes with how Meta handled the Scale AI deal. In that deal, Meta invested in Scale and gave it a valuation of $29 billion. Scale's founder Alexandr Wang agreed to work for Meta and now leads its AI efforts.

• Meta Superintelligence Labs: A look of the challenges ahead

In the end, Nvidia will get Groq's AI accelerator chip technology no matter what you call the deal. The Groq purchase would be Nvidia's largest.

• Nvidia launches Nemotron 3 open models to enable multi-agent systems
• Nvidia's Q3 stellar as data center revenue up 66%