| Constellation Research

Meta's sneak peek 2025 budget: A lot more AI infrastructure spending

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Meta's 2025 budget planning process goes like this: Spend heavily on AI infrastructure and use AI to drive efficiencies so you can plow more money into GPUs.

That peak into 2025 hyperscale budgeting was delivered by Meta CEO Mark Zuckerberg. Zuckerberg was a little more direct about the spending plans on AI, but the other hyperscale giants--Microsoft, Google Cloud and Amazon--have similar plans.

Speaking on a conference call, Zuckerberg was clear that Meta's appetite for GPUs--mostly from Nvidia--will be insatiable.

He said:

"We're training the Llama 4 models on a cluster that is bigger than 100,000 H100s or bigger than anything that I've seen reported for what others are doing."

Zuckerberg added that Meta is early in the budget process, but is targeting the following: AI for efficiency that will partly fund more investment in infrastructure.

1. AI for efficiency. "First, it's clear that there are a lot of new opportunities to use new AI advances to accelerate our core business that should have strong ROI over the next few years, so I think we should invest more there," said Zuckerberg.

Meta's third quarter results were powered by monetization efficiency due to AI. Meta is also boosting engagement and optimizing ad delivery. The company delivered third quarter revenue of $40.59 billion with net income of $15.69 billion, or $6.03 a share. Meta's results were well ahead of Wall Street estimates.

As for the outlook, Meta projected fourth quarter revenue of $45 billion to $48 billion.

CFO Susan Li noted that Meta can boost productivity with AI as it optimizes monetization.

"On the use of AI and employee productivity, it's certainly something that we're very excited about. I don't know if we have anything particularly quantitative that we're sharing right now. I think there are different efficiency opportunities with AI that we've been focused on in terms of where we can reduce costs over time and generate savings through increasing internal productivity in areas like coding."

She said content moderation is another area where AI can boost productivity. Large language models (LLMs) will also improve multiple work streams in general and administrative categories. Li said that Meta also has a headcount opportunities too.

2. Investment in AI infrastructure. "Our AI investments continue to require serious infrastructure, and I expect to continue investing significantly there too," he said.

Specifically, Meta projected 2024 capital expenditures to be $38 billion to $40 billion, updated from the $37 billion to $40 billion range. The company now expects "a significant acceleration in infrastructure expense growth next year."

Zuckerberg, however, noted that infrastructure will become more efficient and that's why Meta has backed the Open Compute Project. He said:

"This stuff is obviously very expensive. When someone figures out a way to run this better, if they can run it 20% more effectively, then that will save us a huge amount of money. And that was sort of the experience that we had with open compute and part of why we are leaning so much into open source here."

Data to Decisions Marketing Transformation Next-Generation Customer Experience Tech Optimization Innovation & Product-led Growth Future of Work Digital Safety, Privacy & Cybersecurity AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Disruptive Technology Chief Information Officer Chief Executive Officer Chief Technology Officer Chief AI Officer Chief Data Officer Chief Analytics Officer Chief Information Security Officer Chief Product Officer

Read more about Meta's sneak peek 2025 budget: A lot more AI infrastructure spending

How to burn down your legacy IT in 10 not-so-easy steps

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

David Giambruno, VP Tivity Health, is the type of person who negotiates his exit package before ever taking a job. Why? He's going to burn your platforms, automate everything possible, cut costs and be hated by everyone at a company except for the CFO.

"I figure out ways to make IT super-efficient, and so I've learned a lot. The biggest thing is having awesome severance and have your lawyer look at it, because everybody's getting angry," he said.

Giambruno has restructured IT operations at Revlon, Tribune Media, Shutterstock and Pitney Bowes. At Constellation Research's Connected Enterprise 2024, Giambruno laid down some truth. "I'm a tech masochist. I never get called by a CTO or CIO. I get called by the CFO or CEO. Generally, if there's some disaster you can fix it," he said.

More from CCE 2024:

Here's a look at the lessons:

Best practices are mediocrity. Most organizations confuse entropy as safety.
You have to burn the platform. "It is about pain and the burning platform. Without that no one ever wants to change," said Giambruno. "Starting a system is the only way to change."
"Spend no money on the old systems. If you're spending money on the old systems you'll never change," he said. "No legacy. Do not waste a second on an old system."

Automation makes everything happen. Automation means more speed and speed always wins.
No multi-cloud. Every cloud you add is about 3x the cost and 5x the security problem.
Run cloud native applications.
Always do proofs of concepts because you'll need to show humans what's possible. "It's really about showing people what's possible because no one ever believes it," he said.
Proofs of concepts are for CEO and CFO primarily, but product people are interested when they realize how much faster you can deliver technologies.

"Nothing runs a computer better than another computer," said Giambruno. Automation means that costs come down and stay down.
"No one will like you. Vendors will hate you because you're taking away huge chunks of money from vendors. I take away huge amounts of money from internal teams too. Then you get a whole new set of vendors and whole new set of processes," he said. "I go from massive chaos to structure."
"Everybody chooses lock in with a vendor. It's cheapest to pick one and then tell them too pound salt when my contract is over," said Giambruno.

Tech Optimization New C-Suite Leadership Chief Executive Officer Chief Financial Officer Chief Information Officer Chief Experience Officer

Read more about How to burn down your legacy IT in 10 not-so-easy steps

Microsoft Q1 strong, Azure revenue growth 33%

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Microsoft reported better-than-expected first quarter earnings as commercial cloud revenue was up 22% and Azure grew 33% from a year ago.

The company reported first quarter net income of $24.7 billion, or $3.30 a share, on revenue of $65.6 billion, up 16% from a year ago.

Wall Street was looking for first quarter earnings of $3.10 a share on revenue of $64.51 billion.

Microsoft CEO Satya Nadella said AI is "expanding our opportunity and winning new customers."

CFO Amy Hood said the company's first quarter execution "delivered a solid start to our fiscal year."

As for the outlook, Microsoft said second-quarter revenue will be between $68.1 billion to $69.1 billion. Wall Street was expecting Microsoft to deliver second-quarter revenue of $69.83 billion.

Nadella said suppliers are late with data center infrastructure and the company won't be able to meet demand.

Key points:

Azure revenue was up 33%.
Microsoft 365 Commercial products and cloud services revenue was up 13% from a year ago.
Microsoft 365 Consumer products and cloud services revenue was up 5%.
LinkedIn revenue was up 10%.
Dynamics 365 revenue growth was up 18%.

Speaking on an earnings conference call, Nadella said:

"At the silicon layer, our new Cobalt 100 VMs are being used by companies like Databricks, Elastic, Siemens, Snowflake, and Synopsys to power their general-purpose workloads at up to 50% better price performance than previous generations. On top of this, we are building out our next-generation AI infrastructure, innovating across the full stack to optimize our fleet for AI workloads."

Nadella added that data centers (DCs) are a constraint. He said:

"We ran into a set of constraints, which are everything because DCs don't get built overnight. So, there is DCs, there is power. And so that's sort of been the short-term constraint. Even in Q2, for example, some of the demand issues we have or our ability to fulfill demand is because of, in fact, external third-party stuff that we leased moving up. So that's the constraints we have. But in the long run, we do need effectively power and we need DCs. And some of these things are more long lead."

Hood addressed capital expenditures.

"Capital expenditures, including finance leases, were $20 billion, in line with expectations, and cash paid for PP&E was $14.9 billion. Roughly half of our cloud and AI-related spend continues to be for long-lived assets that will support monetization over the next 15 years and beyond. The remaining cloud and AI spend is primarily for servers, both CPUs and GPUs, to serve customers based on demand signals."

She said that the capital expenses will pay off over time as capacity catches up to demand.

"In H2, we still expect Azure growth to accelerate from H1 as our capital investments create an increase in available AI capacity to serve more of the growing demand."

Data to Decisions Future of Work Tech Optimization Innovation & Product-led Growth Next-Generation Customer Experience Digital Safety, Privacy & Cybersecurity Microsoft SaaS PaaS IaaS Cloud Digital Transformation Disruptive Technology Enterprise IT Enterprise Acceleration Enterprise Software Next Gen Apps IoT Blockchain CRM ERP CCaaS UCaaS Collaboration Enterprise Service AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Chief Information Officer Chief Technology Officer Chief Information Security Officer Chief Data Officer Chief Executive Officer Chief AI Officer Chief Analytics Officer Chief Product Officer

Read more about Microsoft Q1 strong, Azure revenue growth 33%

Is AI data center buildout a case of irrational exuberance?

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

The cost of artificial intelligence inference and training will fall and enterprises need to question the current groupthink that revolves around a never-ending data center buildout cycle.

Speaking on a panel at Constellation Research's Connected Enterprise conference, Brian Behlendorf, CTO at the Open Wallet Foundation and Chief AI Strategist at The Linux Foundation said "there's a lot of irrational exuberance about the amount of investment that's going to be required both to train models and do inference on them."

Behlendorf said the capacity buildout is going to lead to indigestion.

"I see a lot of enterprises that are cutting other programs, laying off staff and doing everything to conserve capital to be able to collect all the data in the world and build dumb models that they don't really know what they're going to do."

"Yet, the cost of training AI is going to come down dramatically. There are a raft of 10x improvements in training and inference costs, purely in software. We're also finding better structured data ends lead to higher quality models at smaller token sizes."

Behlendorf added that he expects commodity GPU hardware systems to emerge in the next few years. He noted that the idea that the industry is going to need nuclear reactors and an ongoing data center buildout cycle to train large language models is foolhardy.

"A more sober analysis is that you need to build capacity inside your organization at a personnel level and skills level on how to use these technologies and hold on the massive expansion of data centers," he said.

How leaders need to think about AI, genAI

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Artificial intelligence and generative AI will tax leadership, but ultimately raise the bar for decision making.

Those were some of the takeaways from Cassie Kozyrkov, Founder Kozyr LLC, who delivered the keynote at Constellation Research's Connected Enterprise.

Kozyrkov is best known for founding the field of Decision Intelligence and serving as Google’s first Chief Decision Scientist, where she led the charge in Google’s transformation into an AI-first company. She's now an AI advisor for Gucci, NASA, Meta and others.

Here's a look at what leaders need to know about AI.

Every job is going to have some level of disruption due to AI. "Don't think in terms of jobs. Think in terms of what tasks of any given job are most likely to be disrupted," said Kozyrko. "The key thing to understand is that every job has some component that's a little bit repetitive."

Repetitive work is going to be automated. "The repetitive and digitized task is the ideal target for AI automation when there aren't a lot of consequences for messing up performance," she said.

But automating repetitive work can hollow out your bench. "Here's the thing that I think too many people forget. When you hire somebody with no experience at all the work you give them very early on repetitive and easy to check the answer. The perfect task for AI is also the perfect work for your intern or new graduate," said Kozyrko. "A lot more of the junior person's work is going to get cannibalized by more senior folks."

Kozyrko said:

"You should be preparing for what you're going to do with training your future cohort of leaders."

AI washing is trendy. "It is difficult to know what you're buying these days and what to expect--not only of the software systems but the complex collaboration between human and machine," she said.

Here's the test to see if there's AI washing or something like machine learning. "If it's written in Python, it's probably machine learning. If it's written in PowerPoint it's probably AI," said Kozyrko. "Ask questions."

Strategy matters. AI models are just recipes and a human engineer has to think hard about the problem, how to solve it and come up with instructions. "You need to understand the task to come up with those instructions," said Kozyrko. "When humans teach each other, sometimes we use exact instructions. Sometimes we do it another way and teach with examples. Data is just examples."

Leaders prefer examples when control matters. For more complex work, examples matter more. AI will force leaders to raise the bar on performance and be comfortable with change--there's no choice.

AI decisions belong to leaders not subject matter experts. "We have a problem with absentee leadership, where a lot of folks think that AI is the business of the PhD in the world. It is the business of the leader, the decision maker, the domain expert, not the person who's good at mathematics," said Kozyrko. "AI is the product of decision making with some very subjective decisions made by whoever was in charge. The worst thing you can do is think AI is some independent entity that's objective."

Generative AI makes things more complicated for leaders. "Generative AI has more than one right answer and more than one wrong answer," she said. "Test everything and test it in context. Trust nothing you haven't tested and use it carefully."

Kozyrko said:

"It is hard to set criteria for Gen AI and always think in terms of who takes responsibility. That may be more of a limiting factor than some of the technology. So, at the end, is genAI an act of desperation or the frontier of innovation? It's absolutely both."

AI is a genie that can be a friend or foe. "AI will absolutely do is raise the bar for your decision leadership. This is a genie that may grant you a wish, but we know the genie is dangerous," said Kozyrko. "AI will demand more from us in the future. It is absolutely a leadership concern."

Data to Decisions Innovation & Product-led Growth Future of Work Tech Optimization Next-Generation Customer Experience Digital Safety, Privacy & Cybersecurity AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Disruptive Technology Chief Executive Officer Chief Information Officer Chief Technology Officer Chief AI Officer Chief Data Officer Chief Analytics Officer Chief Information Security Officer Chief Product Officer

Read more about How leaders need to think about AI, genAI

AMD's Q3 on target, data center unit revenue growth 122% from a year ago

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

AMD continues to see its revenue surge due to its data center unit, which posted sales growth of 122% from a year ago.

The company reported third quarter net income of $771 million, or 47 cents a share, on revenue of $6.8 billion. Non-GAAP earnings were 92 cents a share.

AMD CEO Dr. Lisa Su said: "record revenue was led by higher sales of EPYC and Instinct data center products and robust demand for our Ryzen PC processors."

Data center revenue was $3.5 billion due to AMD Instinct GPU shipments as well as AMD EPYC CPUs.
PC revenue was $1.9 billion, up 29% from a year ago.
Embedded unit revenue was $927 million, down 25% from a year ago, and gaming revenue fell 69% from a year ago to $462 million.

As for the outlook, AMD projected fourth quarter revenue of $7.5 billion, give or take $300 million.

On a conference call, Su said:

"Data Center GPU revenue ramped as MI300X adoption expanded with cloud, OEM and AI customers. Microsoft and Meta expanded their use of MI 300X accelerators to power their internal workloads in the quarter. Microsoft is now using MI 300X broadly for multiple co-pilot services powered by the family of GPT 4 models."
"Development on our MI400 series based on the CDNA Next architecture is also progressing very well towards a 2026 launch. We have built significant momentum across our data center AI business with deployments increasing across an expanding set of Cloud, Enterprise and AI customers. As a result, we now expect Data Center GPU revenue to exceed $5 billion in 2024, up from $4.5 billion we guided in July and our expectation of $2 billion when we started the year."
"In the Data Center alone, we expect the AI accelerator TAM will grow at more than 60% annually to $500 billion in 2028. To put that in context, this is roughly equivalent to annual sales for the entire semiconductor industry in 2023."
"We feel very good about the market from everything that we see, talking to customers, there's significant investment in trying to build out the infrastructure required across all of the AI workloads."

Data to Decisions Tech Optimization Innovation & Product-led Growth Future of Work Next-Generation Customer Experience Digital Safety, Privacy & Cybersecurity AMD Big Data AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Disruptive Technology Chief Information Officer Chief Technology Officer Chief Information Security Officer Chief Data Officer Chief Executive Officer Chief AI Officer Chief Analytics Officer Chief Product Officer

Read more about AMD's Q3 on target, data center unit revenue growth 122% from a year ago

Google Cloud Q3 revenue up 35% from a year ago, Alphabet results shine

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Alphabet handily topped third-quarter expectations and its Google Cloud saw revenue growth of 35% from a year ago due to generative AI.

The company, which includes Google, Google Cloud and YouTube, reported third quarter net income of $26.3 billion, or $2.12 a share, on revenue of $88.27 billion, up 15% from a year ago.

Wall Street was expecting Alphabet to report third quarter earnings of $1.85 a share on revenue of $86.3 billion.

Google Cloud revenue was $11.4 billion, up 35% from a year ago. Alphabet said Google Cloud saw strength across AI infrastructure, genAI and core services.

CEO Sundar Pichai said the company's "long-term focus and investment in AI" are paying off. Specifically, Google Cloud is driving "deeper product adoption" with existing companies while landing larger deals and new enterprises. How Google Cloud is monetizing AI

Indeed, Google Cloud's operating income is accelerating. The company's third quarter operating income was $1.947 billion, up from $266 million a year ago.

Google's core services remain the cash cow with operating income of $30.86 billion.

Google Cloud has been picking up traction as it builds out services for the AI layer and also drills down into industries.

And more on industries.

Speaking on Alphabet's earnings conference call, CEO Sundar Pichai said Google Cloud's stack is paying off as enterprises leverage AI.

Pichai said Alphabet continues to invest in AI infrastructure--including nuclear power for data centers. Google Cloud is also benefiting from workloads powered by new Nvidia GPUs as well as its own processors for AI workloads. CapEx in the fourth quarter will be similar to the third quarter tally of $13 billion. The largest component of that CapEx was servers, data centers and networking equipment.

Takeaways from the conference call:

Google services are all leveraging Gemini models.
The company has unified teams in AI and machine learnin to move faster. He said the team behind NotebookLM highlights how smaller teams can move faster. "You'll see a rapid pace of innovation," he said.
Google is using AI to generate code that is then reviewed by engineers.
Google is seeing search queries surge due to AI overviews and strong engagement as consumers ask longer and more nuanced questions.
Circle to Search is available on more than 150 million Android devices.
Google Cloud's ability to attract generative AI workloads is landing the company larger deals.
Gemini API calls have grown 14x in the last six months.

On a conference call with analysts, Pichai said the following:

"Customers use our AI platform together with our data platform, BigQuery, because we analyze multimodal data no matter where it is stored with ultra-low latency access to Gemini."
"Each week, Waymo is driving more than 1 million fully autonomous miles and serves over 150,000 paid rides. The first time any AV company has reached this kind of mainstream use."
"On the TPU front, I just spent some time with the teams on the road map ahead. I couldn't be more excited at the forward-looking road map, but all of it allows us to both plan ahead in the future and really drive an optimized architecture for it."
"We are in much more of a virtuous cycle with a lot of velocity in the underlying models. We've had two generations of Gemini model. We are working on the third generation, which is progressing well. And teams internally are now set up much better to consume the underlying model innovation and translate that into innovation within their products. There is aggressive road map ahead for 2025."

Data to Decisions Digital Safety, Privacy & Cybersecurity Innovation & Product-led Growth Tech Optimization Future of Work Next-Generation Customer Experience Google Cloud Google SaaS PaaS IaaS Cloud Digital Transformation Disruptive Technology Enterprise IT Enterprise Acceleration Enterprise Software Next Gen Apps IoT Blockchain CRM ERP CCaaS UCaaS Collaboration Enterprise Service AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Chief Information Officer Chief Technology Officer Chief Information Security Officer Chief Data Officer Chief Executive Officer Chief AI Officer Chief Analytics Officer Chief Product Officer

Read more about Google Cloud Q3 revenue up 35% from a year ago, Alphabet results shine

Takeaways on successful AI, generative AI projects

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Enterprise artificial intelligence projects--generative AI, agentic and everything in between--will often depend on old-school IT management techniques.

Here's a look at some of the takeaways from successful AI projects from a panel at Constellation Research's Connected Enterprise.

It's all about value and use cases. Laurie Wheeler, Chief Operating Officer, Information Services & Technology at MultiCare Health System, said AI projects need to be very clear about use cases and the value expected.

Find a champion. Wheeler, a BT150 inductee, said the projects have done well have featured a "great partnerships with operations." "Having a physician champion was critical to success," said Wheeler.

Don't be wedded to a technology or ideology. Chris Claridge, Chair of Trust Alliance NZ, said stakeholders need to be future oriented but leave technology ideology at the door. "You want people who are open enough about digital identity, fabric ontologies and choose vendors that allow interoperability," said Claridge.

Expectations. Patrick Nicolet, Chairman of Linebreak, said expectations need to be set with AI projects and there's a balance between what can be done and the art of the possible. Wheeler added that you can set expectations to hit a particular metric, but be prepared to be surprised at times.

Standards and quality matters. Wheeler added that champions also help with creating standards and then upholding them as a project scales.

Be prepared to manage fear. Claridge said AI projects often have a tangible fear about job loss attached to them. "That fear causes enormous concerns and fear," he said. "Managing the disruption of this technology is going to be a major issue. It is incredibly disruptive once you start to deploy and scale."

Cultural change. "What's different about AI is that we've been trained to bring answers to questions," said Nicolet. "AI is the opposite. We have to be good at asking the right question and have lots of answers. You really have to shift and that's challenging for any organization."

Claridge agreed and noted that organizations' purpose will be challenged by AI. "Organizations will have to challenge why they exist. AI is going to change the way data moves around and the actual activities of the organization," he said.

Change management. All the panelists said that change management is the secret sauce to AI projects. The communication has to be digestible to end users and people need to be reassured about their jobs. Keep in mind, however, that some of those fears are warranted.

Data to Decisions Innovation & Product-led Growth Future of Work Tech Optimization Next-Generation Customer Experience Digital Safety, Privacy & Cybersecurity AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Disruptive Technology Chief Information Officer Chief Executive Officer Chief Technology Officer Chief AI Officer Chief Data Officer Chief Analytics Officer Chief Information Security Officer Chief Product Officer

Read more about Takeaways on successful AI, generative AI projects

2025 in preview: What Constellation Research’s analysts say

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

In 2025, you'll have to get ready for "knowledge," AI governance will move to the forefront, enterprise software models will be revamped, decision automation will depend on humans in the loop and data strategies will be a pain point.

At Constellation Research's Connected Enterprise conference in Half Moon Bay, the first panel revolved around the first cut of 2025 predictions.

Here's the recap.

Martin Schneider:

Growth strategies and customer journeys will evolve into orchestrated engagements designed for an entire lifecycle.
Revenue plans will be modeled from the ground up by AI.
AI-generated workflows will bring new approaches to customer data that will drive repeatable and scalable predictions.

Chirag Mehta:

Cybersecurity implementations will move toward focusing more on response than prevention.
Generative AI will change the way software is built to make it more secure.
Chief product officers will have all the tools needed to exploit product development for growth. CPOs will no longer be chief backlog officers.

Me:

The enterprise software model will change and alter the way CIOs buy applications. The problem is that revenue models are in flux. Value based models, consumption models and traditional seat models will all be under fire.
Agentic AI orchestration and processes will be critical and a main focus for enterprises in the year ahead.

Andy Thurai:

AI projects will get real budgets and be under more scrutiny for returns on use cases.
AI governance will be critical as enterprises grow concerned about synthetic data.
Here's the problem: AI produces AI that is monitored by AI (see the conflict of interest here?).
2025 will bring more data, more uses cases and more issues. The lawyers will be busy.

Doug Henschen:

Enterprises will realize that they have real data problems to solve before implementing genAI.
"Seventy percent of enterprises in our AI survey aren't seeing the ROI. What these companies have in common is a lack of data, not enough scale and not enough cleanliness," said Henschen.
There will be a barrage of vendor announcements looking to solve these data issues to make data usable for AI.

Liz Miller:

2025 will be the year where enterprises are actually doing what they should be with AI. That means guardrails and a focus on processes. "What we've learned about genAI is that when you automate a really old process with AI you really get a really old result," said Miller.
Enterprise buyers will hear a lot about "knowledge" in 2025, but CxOs shouldn't treat the topic as just another buzzword. Knowledge is about all the accumulated data across the enterprise that drives experiences.

Holger Mueller:

Human capital management will see significant changes. On the people front, employees and gig workers can be doing the same thing. Payroll will become a key sector for innovation. And applications will become more intuitive.
AI will become turbocharged by transactional data.
2025 will be the year of quantum computing (again).

Ray Wang:

Automation is transforming markets and the field will revolve around decisions, not more AI.
"We're going to move the conversation out of AI into agents that can make decisions," said Wang.
Budgets will focus on exponential efficiency due to cost pressures.
Automation success will depend on where you put humans in the process loop. "The number one question for automation is where do you insert the humans," said Wang.
Automation will become less of a concern due to demographics. Most countries won't have enough working people to do the work so automation is necessary.

Data to Decisions Next-Generation Customer Experience Innovation & Product-led Growth Future of Work Tech Optimization Digital Safety, Privacy & Cybersecurity AI GenerativeAI ML Machine Learning LLMs Agentic AI Analytics Automation Disruptive Technology Chief Information Officer Chief Executive Officer Chief Technology Officer Chief AI Officer Chief Data Officer Chief Analytics Officer Chief Information Security Officer Chief Product Officer

Read more about 2025 in preview: What Constellation Research’s analysts say

How Google Public Sector and NASA aim to bring generative AI to aircraft ground traffic control

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Google Public Sector and NASA are training AI models to understand the speech, context and instructions needed to get airplanes from the runway to their gates as efficiently and safely as possible.

Today's airport surface management process revolves around quirky acronyms, aviation vocabulary and human voice traffic that's far from perfect. If NASA Aeronautics Research Institute’s (NARI) partnership with Google Public Sector pans out, the airport surface management process can be augmented with data, speech-to-text instructions and automation.

NASA Aeronautics Research Institute (NARI) is focused on cutting-edge aeronautics research and operational strategies. NARI connects industry, government, and academia to NASA with a focus on autonomous, high-speed, and electric aircraft. "We're the bridge between NASA researchers in aeronautics and the external community, which can be the FAA, other agencies, universities and industry," said Dr. Krishna Kalyanam, Deputy Director, NARI. "NARI was also set up to seed foundational early-stage research that may pan out and turn into a larger project funded by the government."

NARI's priorities include Advanced Air Mobility (AAM), Wildland Fire Initiatives, Shaping Tomorrow's Aviation Systems and providing a collaborative infrastructure for partners to work with NASA.

We caught up with Dr. Kalyanam (right) at the Google Public Sector Summit in Washington, DC to talk about the research project.

The project. Dr. Kalyanam said the goal of the research project is to leverage speech to text models in ground traffic control when planes land and taxi on the tarmac. If the process of getting plans through the airport surface to the gate can be optimized, airlines can improve safety and the cost structure. The research looked into whether voice content over the radio can be turned into taxi instruction with 100% accuracy so it can be absorbed by automation and provide another layer of instructions for pilots.

"You land and you have to get off the concrete. You don't want aircraft on the runway and the sooner you can get an aircraft to its destination, the more planes you can get on the runway," said Dr. Kalyanam. "As soon as you land, you're getting instructions to the gate assigned to you by your dispatcher. All instructions are provided by the ground controller. It's 'take this route. Turn here. And here.'"

Challenges. The biggest challenge with the project, according to Dr. Kalyanam, was that voice traffic between pilots and the control tower has its own vocabulary as well as poor radio quality.

"Say you're running into some bad weather and need instructions to the gate. Today that's full end-to-end speech. The information could be augmented by text, visual and other inputs to go along with voice that can be converted to a route that's communicated digitally," said Dr. Kalyanam. "Once digital it can be displayed on a map or directly ingested into route planning."

Another challenge is that instructions to pilots have a unique vocabulary including terms like Roger and Wilco that humans can easily fill in gaps when interpreting data. Models need to be trained on voice traffic over air to pick up this vocabulary.

More from Google Public Sector Summit:

The goal. By digitizing the voice traffic over radio, directions can be given via moving maps, text, and color codes. That data can also be used to optimize routes and improve efficiency. "Once you digitize the information you have all this information in one place that can be optimized," said Dr. Kalyanam. "There are 100 tasks that are needed between the time the plane lands, people get off and the plane is ready to take off again."

Dr. Kalyanam said this research could also apply to autonomous aircraft and refueling. "The traditional processes are mostly human-centric," he said. "Some of these things can be automated, but at the least you can make it easier for humans to perform tasks.”

He added that the motivation of the research is to provide a secondary source of information for the pilots.

Training models. Google Public Sector used multiple models for training, but training was done on a minimum data set of 10 hours of voice instructions. Google's base models were already trained on general English conversations but had to be customized by vocabulary and use case. The models would pick up voice instructions, transcribe them according to ground control's acronyms and vocabulary and create digitized instructions.

"It's almost like learning a new language," said Dr. Kalyanam. "There are words that we will never use in English because they mean completely different things. You need to get the right context. If you hear “Dealt” it most probably means “Delta” with the ‘-ah’ sound clipped. Sometimes you can’t hear parts of what is being said. You're training models to be as perfect as they can be in an imperfect environment."

Google Public Sector and NASA worked with retired controllers to verify the ground truth as well as the voice commands and how the models performed. "The goal was to capture the taxi instruction with 100% accuracy," said Dr. Kalyanam. "There may be a conversation, but you want to be able to know that the pilot can't turn left from Charlie to Lima and then use that information. There's local knowledge about the airport layout that can be used to fix errors.”

Complicating the training effort is the reality that every airport is different, and the models will need to know local context—say the differences between the airport in Dallas vs. Tampa. It’s possible that models will need more fine tuning based on location of the airport. This fine-tuning will be even more important if this research is applied to international airports.

Working with Google Public Sector. Dr. Kalyanam said partnering with Google Public Sector made sense given Google's experience in AI, speech-to-text use cases and mapping. "We had some internal stuff we developed over the years, but the speech-to-text expertise was with Google," said Dr. Kalyanam. "Google has the models and has done the research."

Dr. Kalyanam added that Google Public Sector also had the engineers available to test multiple models and configurations for the audio. "Not one single model works best," said Dr. Kalyanam. "It took a lot of experimentation. This is custom engineering work. It's a good partnership since we don't have access to what's inside the box, but we can provide feedback so Google can build something. We also have retired controllers and pilots for model validation."

Metrics. Although it's early in the research process, Dr. Kalyanam said time saved and reduced mishaps will be core metrics. "If you end up in the wrong place it's a lot of time wasted because aircraft normally do not go in reverse," he said. "There's a lot of opportunity with digitized data. If you didn't make a turn, automation can alert you and give you new instructions. I think this process can be made simpler and hopefully less prone to error."

What's next? NASA and Google Public Sector are looking to publish their research and work with the FAA and the aviation industry. "There's a lot of interest in this research," said Dr. Kalyanam. "This is exploratory research, so we are ready to accept some failures. We are trying to prove this concept and maybe we'll simulate it in one airport and see how it adapts. We do the research, crunch the numbers and work with the FAA and industry to mature the technology for use."

Data to Decisions Next-Generation Customer Experience Digital Safety, Privacy & Cybersecurity Innovation & Product-led Growth Tech Optimization Future of Work Google Cloud Google SaaS PaaS IaaS Cloud Digital Transformation Disruptive Technology Enterprise IT Enterprise Acceleration Enterprise Software Next Gen Apps IoT Blockchain CRM ERP CCaaS UCaaS Collaboration Enterprise Service ML Machine Learning LLMs Agentic AI Generative AI AI Analytics Automation business Marketing finance Healthcare Customer Service Content Management Chief Information Officer Chief Technology Officer Chief Information Security Officer Chief Data Officer Chief Executive Officer Chief AI Officer Chief Analytics Officer Chief Product Officer

Read more about How Google Public Sector and NASA aim to bring generative AI to aircraft ground traffic control

Subscribe to

Meta's sneak peek 2025 budget: A lot more AI infrastructure spending

Meta's sneak peek 2025 budget: A lot more AI infrastructure spending

How to burn down your legacy IT in 10 not-so-easy steps

How to burn down your legacy IT in 10 not-so-easy steps

Microsoft Q1 strong, Azure revenue growth 33%

Microsoft Q1 strong, Azure revenue growth 33%

Is AI data center buildout a case of irrational exuberance?

Is AI data center buildout a case of irrational exuberance?

How leaders need to think about AI, genAI

How leaders need to think about AI, genAI

AMD's Q3 on target, data center unit revenue growth 122% from a year ago

AMD's Q3 on target, data center unit revenue growth 122% from a year ago

Google Cloud Q3 revenue up 35% from a year ago, Alphabet results shine

Google Cloud Q3 revenue up 35% from a year ago, Alphabet results shine

Takeaways on successful AI, generative AI projects

Takeaways on successful AI, generative AI projects

2025 in preview: What Constellation Research’s analysts say

2025 in preview: What Constellation Research’s analysts say

How Google Public Sector and NASA aim to bring generative AI to aircraft ground traffic control

How Google Public Sector and NASA aim to bring generative AI to aircraft ground traffic control

Research

Analyst Services

Videos

Communities

Events

Insights Live

Results

Meta's sneak peek 2025 budget: A lot more AI infrastructure spending

How to burn down your legacy IT in 10 not-so-easy steps

Microsoft Q1 strong, Azure revenue growth 33%

Is AI data center buildout a case of irrational exuberance?

How leaders need to think about AI, genAI

AMD's Q3 on target, data center unit revenue growth 122% from a year ago

Google Cloud Q3 revenue up 35% from a year ago, Alphabet results shine

Takeaways on successful AI, generative AI projects

2025 in preview: What Constellation Research’s analysts say

How Google Public Sector and NASA aim to bring generative AI to aircraft ground traffic control