And here we are today, each of us subjected to a torrent of news and information, texts and emails and social media feeds, conversations and team chats, news articles and opinions, stock prices and analysis, numbers upon numbers, words upon words, and still, no loom. And too often, the sources of news we’ve counted on now tell us what is happening, but not always what it means, or why we should care, or who we should trust. As the Chief Communications Officer for Microsoft for more than 15 years, I’m both a victim of this deluge of facts, and a contributor to it. I regularly read three or four daily newspapers, mostly get through both the New Yorker and the Economist, have news alerts, scan forums and aggregator sites for tech news, listen to probably ten podcasts and at the end of the day often wonder… what was true? Through this lens, welcome to the first edition of Microsoft Signal, a magazine for senior business leaders interested in hearing directly from Microsoft and from each other. How is technology changing the way we work? What are best practices, by industry, on how to thrive? Who is doing interesting work? What does the future hold? In this changing landscape of media, each of us has choices about where we go for news, information, context. We hope you’ll appreciate the chance to hear from us, in areas where we have expertise, and from your peers on the topics most important to them. Enjoy this first issue! And let us know what else you might want to see.

Frank X. Shaw,
Chief Communications Officer, Microsoft

When I prepare for customer meetings today, I go to Researcher for Copilot. It’s like having a really smart analyst who composes everything from the web, from SharePoint, from my CRM database, and brings it all together so that I’m briefed. Instead of sending five mails, getting six documents, and assembling all of that information, it’s now one click away. Same thing with data analysis. It will mean a significant change to what I describe as the new production function for knowledge work. Whether it’s tech and software, or health care, financial services and even the back-office operations of any manufacturing company, production is going to be fundamentally redone, rethought. It’s going to be more and more efficient. The mobile and cloud revolutions were big, but they were incremental. We’re doing things 10x faster now. That’s convenient but it does require significant change. And talking about that, Marcus, I love your new book, A Platform Mindset, and the way you framed it. You’ve seen this from a variety of vantage points – from Microsoft to Google, to Stone. What’s your take on what’s happening and what is the big takeaway for you?

MF: I wrote the book as an examination of how culture can leverage the best skills in an organization to foster innovation. Platforms have a multiplier effect because they can help us imagine new scenarios and new businesses and then scale them quickly. The title was inspired by the growth mindset you’ve cultivated at Microsoft. How do you create this standardization in the platform so that teams can focus on creative and innovative solutions?

SN: Your perspective is beautifully said. More and more of our work is driven by digital platforms. An organization improves the more it integrates continuous improvement into the platform. The bottom line is you need a leadership mindset that cultivates this. Investment decisions are not about any single feature. It is about a platform that enables you to build features faster. That’s the biggest change. More and more of our work is being driven by these digital platforms. One thing you learn very quickly is the integrative effects; you want to make continuous improvements in the platform to leverage your investment.

MF: I write about an engineering culture that I like to call fearless execution. I want to empower the engineers to be able to just write code and put the code in production without being afraid that they will be singled out for problems down the road. We have such rigorous checks and balances across the stack, throughout the platform. Problems will happen but I want to encourage confidence individually and as a team. I’d love your take on this.

SN: I love all those terms you use. In fact, I’ll add one more, which is toil. As leaders, we are responsible for finding the toil of people, our engineers. It’s frustrating if they are having a tough time making code changes and getting them deployed because of all the manual processes. The key thing is to standardize, standardize, standardize; automate, automate, automate. Build processes that you can really depend on so that when you push the execute button there will be lots of checks, there will be lots of tests. And even when you deploy, you will deploy to a small percentage of users. It will automatically revert if it doesn’t work, and it will give you a notification to fix something. There’s so much rigor that has been applied to standardization and automation to help make sure that you are not going through so much toil. If we can get that, then we’ve enabled our people to produce more and more innovation. We can focus on how we are going to get better. Something obviously will go wrong. That’s where you embrace the red. You want to see scorecards that are mostly red, so that you can go back and say, what should be standardized? What should we build that is robust? How should we make sure our flighting system or experimentation system allows us to test things on small sizes before we have a massive blast radius?

MF: I love that. I want to ask you about innovation. We all live in this world of tight budgets, and we need to do the next thing, the next feature that our customers are demanding. On the other hand, you also have to plan for the future. How do you balance that?

SN: The constraints are very real. There are time-to-market constraints, cost constraints, and many others. But leadership is all about picking, deciding. It’s not about doing all things. It’s about making choices on what to do. What will add value for customers and what will competitively differentiate. Start there, but then really use the platform effects to your advantage. One of the classic things we all fall into is the trap of, “Oh no, my competition is doing x, now I’ve got to do x”, so therefore I need incremental revenue. You’ve got to reframe that: in order to do x to match a competitor or customer expectation, how can I do that better than anybody else because of the platform investments I’ve already made. Build that strength. And so, we have to become very good at constraint solving, because, of course, all of us can be great if there are no constraints. But the reality of leadership and business is all about managing constraints and yet remaining competitive.

MF: It’s like using platforms and the leverage that you get from platforms to create efficiencies that you can then reinvest to innovate; to do things in a different, more productive and scalable way.

SN: Exactly. You had a good reference in your book. You called it the bicycle for the team. You want to talk a little bit about that?

MF: Yeah, the idea is that there is always this time-to-market constraint. But think of a long race, a marathon. If I start running in front of you the fastest I can, you should probably stop to build a bicycle. It’s worth “wasting” the time needed to build long-lasting bikes – the races are long, and those bikes will help us win not only the current races, but also the ones we cannot yet foresee. I’m convinced technology companies need generic, reusable platforms to have a competitive edge.

SN: It’s a nice metaphor. I know Steve Jobs had that metaphor of computers being like bicycles for the mind. And this is another way to say it. Platforms are bicycles for teams to get ahead. That’s a great way to describe the leverage of a platform mindset.

MF: This requires a cultural change. How would you go about doing that, and especially in a large organization?

SN: Ultimately, I think a lot of leadership challenges come down to not being clear. So having that ability to bring teams together and drive clarity around what the ultimate goal is, is critical. It’s not about any one thing. It’s about ultimately doing something of significant impact and value to customers, and being able to get back to framing that with clarity, creating the capability that is needed for it, which is, how do you build? In order to win that race, let’s make the right investments and really constraint solving for what ultimately is the winning play and that building the right expertise in the team, the right platforms in the team to go after it, and then bringing, quite frankly, culture and the energy to your point. And so, ultimately, I think of this as getting clarity on the concept, having the capability to go after that concept with this platform leverage, and then I would say the culture that allows you to build that capability to go after the concept. I always go back to those three Cs.

MF: Thank you so much, Satya. I think this was great.

SN: Fantastic. I’m so glad you wrote this book. And I think this would be a very useful thing for a lot of people, because, quite frankly, as both computing and AI becomes so much prevalent in all walks of life, in all organizations, I think this idea of really empowering people with the latest tools and then having a platform mindset, I think will allow us to drive, ultimately, what is our collective goal of driving economic value and growth all around the world and so I’m really looking forward to it.

Petty and the Vodafone team split their application strategy into three components: internal productivity, external interfaces and building generative AI into their products. The third of these is a work in progress, but the first two are already well-advanced and have generated significant learnings. “We find thinking about it across those three pillars is a very effective way to stop being overwhelmed,” says Petty.

The first pillar, he says, involved “harnessing the benefits of generative AI to get rid of the drudgery of routine administrative tasks and to help accelerate processes.” This involved deploying Microsoft Copilot to 50,000 users across the organization. “I have to be honest, this was a bit of a leap of faith, but after six months we really saw a return economically,” he says.

Petty cites the use of Copilot across Vodafone as a prime example of generative AI in action, along with more specific tools such as Power BI for business planning and GitHub Copilot for software developers. While the investment in generative AI could be seen as a risk, Petty says this was mitigated by the speed of the returns. “With a normal IT project you can spend a year deploying something and then a year training people and then eventually in the third year you get some value,” he says. “We’ve seen with generative AI, you deploy very quickly, people get up to speed very quickly and you start getting value much faster than you would in a traditional IT project.”

We asked Petty what he learned through rolling out AI internally and in external interfaces – these are the tips he gave us for anyone looking to do the same…

1. Share the power

Petty believes that the success the company has had in increasing productivity via generative AI is down to the efforts it made in training. “We have champions in every business unit that show people how to use GenAI technology,” he says. “I do it myself. I show how I use Copilot to prepare for meetings, to create external briefings, to monitor all the things that I need to track.” He highlights the danger of assumed knowledge. “Don’t presume that because you’ve given your employees a new technology tool that they’re going to immediately benefit from it,” he says. “You’ve got to make sure that you’re educating them.”

2. Go all in

Today the use of generative AI is seen as a positive by most in the company. Petty also thinks Vodafone benefited from making the tools available companywide. “We didn’t want to cherry pick and give it to a few people in, say, finance or legal, everyone needed these tools,” he says. “It’s become a bit competitive internally,” he adds. “We saw cases of different business units comparing how much value they were getting from [GenAI] versus other teams.”

3. Supercharge your sales

“In every company you have a basic conundrum: The more salespeople you have, the more you’re going to sell, but if those people aren’t working efficiently, you may not be able to justify their costs,” says Petty. A huge percentage of a Vodafone salesperson’s time used to be spent generating proposals…

4. Support your support team

“Our non-GenAI chatbot, TOBI, was OK at answering pretty standard questions, such as ‘What’s the price of roaming in the United States?’, but not very good at more difficult questions or longer conversations,” says Petty…

6. Create an ocean of data

Petty believes that the success Vodafone has had with generative AI is down to the quality of its “data ocean” but admits that creating it wasn’t easy…

7. Clean up your act

To combat the problems of confusing the AI, the team invested in lifecycle management, building an application that could encourage individuals to clean up their data act…

8. Classify your documents

Securing sensitive information is another issue – the AI needs to know what is appropriate for certain users as it might otherwise surface sensitive financial records or classified information. “We have a four-level system in Vodafone that we use Office 365 for,” Petty says. “You tag every document. C1 means public, C2 means generally available, C3 means can’t be shared outside the company and C4 means confidential. As long as you tag your document correctly, Microsoft’s GenAI will apply the right rules and the right controls to sit around that.”

9. Act now

While Petty warns about getting sucked into “the hype cycle” around generative AI, he does believe that businesses yet to engage with the tools risk being left behind. “Generative AI is going to transform every industry in the same way that the internet and mobile apps did,” he says. “It is probably not going to happen as quickly as the hype cycle says, it could take three to seven years, but in that window, the way businesses run will fundamentally change.”

Petty believes there is a window of opportunity for companies to capitalize on the adoption process. “The trick for leadership is to work out where AI is going to add the most value soonest in their industry and unlock that value quicker than their competitors,” he says. This will be achieved by businesses limiting focus to a few areas that have the potential to scale. “There’s so much hype about GenAI that I think a lot of businesses are not quite sure how to get to scale,” Petty says, adding that he’s seen a lot of companies that are running as many as 500 proofs of concept (POCs) across their business. “I think it’s important to ask how many use cases you have with 10 million users or 20 million users,” he says. “I’d argue that without that you’re not really extracting value from generative AI. You need to get to scale in a small number of cases to unlock value, not run hundreds of POCs that never really deliver into operational benefit.”

10. Stick to what you’re good at

When asked for his final piece of advice, Petty talks again about the dangers of companies getting drawn into sinking huge amounts of money into the technical infrastructure instead of concentrating on their own area of expertise. “My fear for many companies is that they focus in the wrong place,” he says. “They talk about how many processing chips they’re going to buy. They talk about training their own large language models or doing their own fine-tuning. I think they’re making a mistake; they should really focus on the application layer. That’s where the value comes from. You can rent all the infrastructure you need from Microsoft, Google, Oracle or the other companies spending a large amount of CapEx to build that. If you can create, say, a 15 percent productivity increase over your nearest competitor by using Copilot, then you can grow faster than them. It’s a huge competitive advantage.”

Yet when given the opportunity, the Nou Sardenya can generate an atmosphere like no other – and right now, three minutes into the game against Real Racing Club, the fans of Europa women’s team are celebrating a corner kick with surprising fervor. As winger Júlia Gómez prepares to take the kick, she crosses her arms to signal her intentions to her teammates. The majority of the team runs into the box, drawing opposition players with them. But the ball is rolled to the near side where captain Pili Porta meets it, arching a perfect shot into the top corner. The crowd erupts, its celebrations rattling the windows of the apartment block that towers over the south stand.

Fans of Europa haven’t had much to cheer about in recent years. The men’s team was one of the founding members of La Liga, the highest division in Spanish soccer, but has since slipped to the fourth tier of the country’s league system, overtaken by clubs with bigger stadiums and budgets. Last season, after a disastrous run of results, the women’s team was relegated from the second tier in which it has played for most seasons since it was founded in 2001, to the third – raising real fears about its future. “In women’s football in particular, relegation can mean disaster,” says Europa’s coach María Victoria Haces, known by all at Nou Sardenya as Nany. She explains that clubs can get locked in a downward spiral, with the loss of revenue and prestige following demotion leading to a mass exodus of players and staff, which can often lead to further relegations. “You see so many relegated teams disappear within a year or two,” she says. Assistant coach Lucía Martínez agrees. “I think the relegation made us think about what we were doing and the things we need to do better,” she adds. “It also makes you look for new ways of doing things.”

This season, Europa has indeed tried something different – and it seems to be working. As expected, the team lost some key players and members of the coaching staff after relegation. Yet rather than sinking to the foot of the table, as many had feared, when Signal visits the stadium in March 2025 Europa is flying. With seven games of the season to go, the women’s team sits near the top of the table, locked in a three-way battle for promotion. One of the new things the club has embraced is the use of AI. And the goal that sparked the window-rattling celebrations may well be the result of the new approach.

Backing the underdogs

The arrival of AI in sports has threatened to widen the gulf between the elite teams and the rest of the pack further. Clubs like FC Barcelona are investing huge sums in AI, and they seem to be paying dividends. One insider says that following machine learning analysis, the starting positions of the FC Barcelona men’s defensive line was moved by the tiniest of margins in the run-up to the 2024/2025 season in the hope of winning more free kicks by catching opposition players offside. In the first half of the season, the team won 201 free kicks this way, more than twice as many as any other major club in Europe.

But an initiative led by Founderz, the new sponsors of Europa’s women’s team, is trying to prove that you don’t need to spend millions to get good results. Founderz is an online platform that uses AI to bring the learning, collaboration and networking opportunities available at the world’s best business schools to the online world. The company was initially approached by Europa to simply sponsor the team’s shirts, but the joint CEOs Pau Garcia-Milà and Anna Cejudo soon sensed an opportunity for a deeper collaboration. “We became fans,” says Cejudo, who played soccer herself as a teenager. “And like all fans you want to do what you can to help the team. Europa doesn’t have huge resources and we thought AI could bridge that gap.”

“The goal is to use AI tools that are available to everyone to provide the team with the insights they need to be competitive,” adds Garcia-Milà. “Can we use these tools, that weren’t available two or three years ago, to affect real change? Can we help?” Their efforts to assist the team began with Founderz building an AI model that the coaches could interrogate using Copilot. “There’s a lot of data in football at every level,” says Garcia-Milà. “We fed in match reports, statistics on running speed, how goals were scored, video… We showed the model to Nany and the team at the start of the season and told them how to interrogate it via Copilot.” The team could ask the AI questions such as: Who is the fastest player in the team we are due to play next weekend? What is their starting line-up likely to be? Where are we losing possession? Where do the moves that lead to our goals usually start? “The coaches seemed really impressed, but they didn’t use it,” says Garcia-Milà with a laugh. “I’d ask them ‘How are you getting on?’ and they’d say ‘Fine’, but I could see exactly how many queries the model had fielded. And it was none.”

Rebooting soccer

The reason for this initial lack of engagement with Copilot was time. Alongside coaching the women’s first team, Nany is also sports director at Europa, and a lot of the staff have other jobs to supplement their modest income from soccer. “We didn’t want to force things,” says Garcia-Milà. “We thought this could be useful, but we know the realities of the workload of everyone in the club. So we left it.” The project seemed to be over before it had begun – but as in all good sporting stories, salvation came just in the nick of time. A group of students who had been using the Founderz platform reached out to Garcia-Milà; they had heard about the company’s involvement in Europa and wanted in. Together they created a new way of working in which the students became researchers, interrogating the data using Copilot to produce reports to help the coaches prepare for matches. “They knew the Microsoft tools, they love football and Nany was super open to the idea of them being involved,” says Garcia-Milà. “It was the perfect match.”

Moneyball meets Ted Lasso

Sitting at the table in Europa’s office it is easy to see why Endika Alonso, Angela Blanco, Valeria Coto, Álvaro Rivera and Daniela Pérez-Pasten make such a good team. They are from different countries – Spain, Bolivia, Costa Rica, Chile and Mexico respectively – but have a shared passion for soccer and the potential of the technology, and talk about both with an infectious enthusiasm. “If we do this right, we can prove that these types of tools can help a team without all the resources in the world to compete like a really big club,” says Rivera. “At Europa’s level you can really see the difference between clubs that have money and those that don’t,” adds Coto. “I believe that this can level the playing field.”

Every week the research team receives a raft of data from the coaching staff and external sources. It can include everything from voice or text notes from Nany and the team with ideas and observations to GPS data from trackers in the players’ kits and videos of opposition teams. This is all fed into the model and the researchers then interrogate that data in different ways to create an easily understandable report for the time-poor coaches.

“Like football it is a mixture of art and science,” says Rivera. “The art is learning how to ask the right question of Copilot, the prompt that yields the most valuable insights.” The researchers then compile the best findings, such as profiles of the opposition, suggested line-ups and possible areas of advantage, into the report. What the coaches do with that information is totally up to them. “This project is Moneyball meets Ted Lasso,” says Garcia-Milà, referencing the 2011 biopic starring Brad Pitt and the feel-good soccer comedy featuring Jason Sudeikis in the title role. The former, based on a nonfiction book by Michael Lewis, looked at how a baseball team used statistics to assemble a competitive team on a limited budget while the latter was all about the importance of communication in turning a group of individuals into a successful team. “You need the human interpretation of that information to bring it to life. That’s Nany’s genius.”

Pérez-Pasten says that the reports are improving all the time, thanks to feedback from the coaches. “You keep going, but dig deeper,” she says. “The more we learn, the better we can get.” Blanco agrees. “A football team doesn’t come together overnight, we need to work at it,” she says. “Every week the reports are getting better because we are getting more data and we are learning what we can do with it.”

One of the areas the team decided to focus on was set pieces: dead-ball situations such as goal kicks, free kicks and corners. Speaking the day before Europa scored from just such a situation, the team hinted at what was to come. “This week we’ve done a lot of work on player positioning,” says Alonso. “The starting positions from corners, what could we do to take the opposition team by surprise.” Lucía Martínez is the assistant coach responsible for set pieces. “AI can make things faster,” she says. “I watched a few videos and shared my observations with the research team. They then applied these insights across a larger sample – 40, 50, 60 videos – to verify patterns. That kind of analysis would take me hours, and I don’t have that kind of time.”

Against Real Racing Club, the system seemed to work perfectly. After Porta’s well-worked goal the players ran to the bench to celebrate with Martínez, suggesting that the successful corner routine had come straight off the training ground. After the game, neither the coaches nor the researchers would say whether AI insight was behind the goal, with inquiries met with a series of wry smiles. Whoever or whatever was responsible, the goal was the foundation for a 2-1 victory. With Europa’s two nearest competitors drawing on the same day, the win took them to the top of the table.

A game of two halves

Nobody at Europa is suggesting that the AI system is some sort of cheat code that will allow any team with access to the technology the ability to suddenly win every match. “We have to be honest that football is played by 11 people versus 11 on the pitch,” says coach Nany. “The teams with the most money can afford the best players and the players will ultimately make the difference, but when the teams are close then small things can have an impact. I hope that soon the AI will start to show us things that the eye cannot see. To see patterns and reveal tactics that we cannot. We are not there yet, but we are beginning to see marginal gains.”

Gains are also being made off the pitch. As part of its involvement in the club, Founderz made scholarships available to all the players on the women’s team. The scholarships allow the team, who play on a semi-professional basis, to study several subjects, including AI, on the Founderz platform. Two players embracing the opportunity are Adriana Manau and Aina Ortiz. Manau juggles her work as an architect with playing as striker for Europa. She also plays beach soccer and represented Spain at the sport’s recent World Cup. “I would not change my life, but it is very busy,” she says. “Every second is accounted for.”

In a bid to claw back some precious time, she has started using Copilot in her day job, which involves sending out tenders for public building projects. “I have learned a lot with Founderz, and I am applying it every day. It has helped me be more efficient at writing reviews and tenders. I have such little time, so it is perfect.”

Ortiz, meanwhile, was just a month into her career at Europa when she was badly injured. Damaging her anterior cruciate ligament (ACL) and meniscus, she faced a year on the sidelines and, having already studied sports science, jumped at the opportunity to expand her knowledge with Founderz. “These types of courses can help you grow,” she says. “It is incredible what you can learn on this platform.” Ortiz now hopes to use her newfound knowledge to help others avoid the types of injury she is suffering from. ACL injuries account for nearly a third of playing time lost to injury in soccer, with female players thought to be up to six times more at risk than their male counterparts. “I think AI, with the access to the right sort of data, can help avoid these injuries,” says Ortiz. “My injury happened in the 14th minute, I wasn’t tired, I was warmed up, it wasn’t a foul. There is a video of my injury and there will be hundreds of other videos. I hope that AI can help find the connection and we can avoid future injuries. That for me is as exciting as any goal.”

With seven games left to secure promotion, the researchers’ efforts are focused on the performances on the pitch, but they too are looking at other ways AI can have an impact. “I think this can be bigger,” says Coto, to enthusiastic nods from her fellow students. “We are all studying business alongside AI and football is a business. Once the season is over, we are going to look at how AI can help grow the club. We are hoping to launch an agency that not only looks at performance but also at how AI can help a club sell more tickets and attract more sponsors. We are very optimistic that the AI will help a club find more fans, to research the perfect moment to sell more tickets and how to target sponsors. A full stadium can only help the performances on the field and the finances off it.”

Founderz is thinking of the bigger picture too. “There’s a full circle of things that you can do with AI,” says Cejudo. “For example, AI can help with fan engagement. AI can predict if it is going to be a cold day, so you can stock the bar with more coffee. It can help you sell more jerseys online. All of this adds to the income of the club. It can help them grow.”

“We know there is a gap in terms of technology and money,” concludes Garcia-Milà. “But you can think of it in a different way. So-called ‘bigger clubs’ are investing millions in bespoke technology. But Microsoft has invested billions in this technology. All we need to do is unlock it.” At the start of April, the league pauses for its spring break to mark Easter celebrations. Europa remains top going into the break and hopes are high that it can secure a return to the second division. Regardless of how the season ends, the team has at least one new set of supporters. “We don’t see an expiration date,” says Cejudo. “When we commit, we commit fully. We are all in. We can’t wait to see where this can go.”

Tackling the corn maze

Topology, the discipline that started everything, focuses on geometric configurations that are not altered when they are bent or stretched, which means that structures can be disturbed without being transformed. Topology could allow for a new future for the notoriously fragile qubits, one that could see them withstand being disturbed, say, to be looked at or measured – and maybe even harnessed to power a new type of chip.

Unlike a regular computer bit, which can hold information as either a 0 or 1, a qubit – usually made from subatomic particles – can hold both values at once, which means it can process exponentially more information, exponentially faster. Think about it as the difference between exploring a corn maze by following each path one at a time then retracing your steps to start the next path versus exploring every path at once.

This ability means quantum computers could potentially find solutions for problems of a global scale such as climate change. That potential to make the world better was why Nayak and others at that dinner couldn’t let go of the idea. With his in-flight plan in hand, Nayak connected with some colleagues and in 2004 they wrote a paper detailing their approach.

Another attendee at the fateful Harvard dinner was renowned mathematician and researcher Michael Freedman, who had been exploring quantum topology and physics as part of Microsoft Research’s Theory Group for several years. After the dinner, he suggested to Microsoft executive Craig Mundie that the company should invest in researching topological qubits. After a few conversations, Mundie not only greenlit the idea, but also told Freedman that he would now have to be both a mathematician and a program manager – and that he had better start assembling his team.

Freedman wasted no time in hiring experts in math, computer science and physics, some of whom had attended the Harvard dinner. One of them was Nayak, who joined as a senior researcher of quantum hardware. In 2005, Microsoft established its first quantum research lab, Station Q, and the team got to work. Their first job? To find something that may not even exist.

Straight up the mountain

Their idea that a qubit could be rendered more stable by being topologically protected relied on the existence of exotic particles called Majorana fermions, which were still only theoretical, having never been seen or made.

So why would an organization known for practical tools such as Windows and Outlook take a risk on a theoretical particle that was not even guaranteed to help their mission if they did find it? For Mundie, the answer was obvious: “To me, the reason to do this was that it was going to change computing.”

Peter Lee, head of Microsoft Research, believes that the company’s willingness to invest decades in quantum was rooted in its fundamental vision. “Microsoft isn’t only doing the here and now but is creating the conditions for success and growth in the future,” he says. “One of the things that means is that we actually do invest in things that will define that future.”

But while the will was strong, the challenge was immense. Not only did the team have to prove the existence of the Majorana and create the architecture and materials to house and control topological qubits, but they also had to develop the software that would someday run on the new quantum computers. Mundie has described the multidisciplinary effort as “the most complex engineering task that humans have ever undertaken.”

You can think of the conventional route to quantum computing as akin to setting off on an easy hiking trail, making steady progress for a while but eventually reaching a cliff, which you always knew was there, that completely blocks progress. Microsoft, on the other hand, decided to labor straight up a seemingly unscalable mountain.

The qubit hunters

Freedman’s team at Station Q partnered closely with academics who could set up experiments to try to detect Majoranas. Having the experimental portion of the work carried out externally made sense at the time because it can take years to build a lab, which would have been a massive disruption to the project. There were other benefits, too. “We would get a lot of exciting science happening at universities, and students were being trained who we could eventually hire,” Nayak says. “We viewed it as a conveyor belt of talent toward Microsoft.”

In 2012, there were the first glimmers that the unconventional path just might be the right one: a team at Delft University in the Netherlands that was working with Microsoft had detected evidence of Majorana quasiparticles. These signals did not yet demonstrate the protected qubits of theory, but they did excite the physics community. Reflecting on the discovery in 2014, Lee said: “It’s not definitive proof, but very strong evidence, and several other experimental physics groups around the world have since come up with similar results in their own independent experiments.”

By late 2016, it was time to bring the experimental arm of the topological qubit-hunting journey in-house. Microsoft entrusted the project to the leadership of engineer Jason Zander, who was instrumental in building the first enterprise-scale Azure cloud for Microsoft.

“Once we became firmly convinced that this stuff would work, the level of engineering precision and scale got very serious and researchers and research leaders like me, we’re just not going to be that reliable taking something like that on,” Lee says. “I mean, if you’re building the Large Hadron Collider, you don’t have a bunch of theoretical physicists doing that. You have real construction engineers and people who know how to oversee an army of people who do that.”

Dodging roadblocks

To create conditions to prompt the formation of Majorana fermions and support topological states, Microsoft needed to combine the properties of a semiconductor with superconductivity.

But which semiconductor material to use? The team had a couple of ideas and had to choose one because working with two different semiconductors would mean creating two complete sets of tools to avoid contamination. Should they go with indium arsenide or indium antimonide?

“This was a decision with imperfect information,” Nayak says. “We said, ‘Let’s pick one and fail fast.’”

Lee says that this choice highlights a fundamental difference between how research is done at Microsoft versus at a university, where researchers typically have more freedom to explore multiple ideas – even if that isn’t always conducive to rapid progress. “At some point things get expensive enough where you just want to force a level of focus,” Lee explains. “Yes, there are some uncertainties, but this is the path that we’ve chosen, and we’re going to devote all of our time and attention to do this.”

Indium arsenide won out. The team concentrated all their efforts on seeing whether it could be used to create nanowires a thousand times smaller than a human hair – and if these wires, when placed in the right magnetic field at the right temperature, could support Majorana fermions at their ends.

Experiment followed experiment as they looked to coax Majoranas into existence and harness them. “We were just smarter every day than the previous day,” Nayak says. “And we were making progress. Not necessarily linearly, but we weren’t stuck at an impasse for any period of time. We would hit roadblocks. We’d figure out a way around them.”

In 2021 came an announcement that seemed like it could be more than a roadblock – it could be the end of the road entirely. A 2018 paper by researchers at Delft University of Technology’s Microsoft Quantum Lab that had shown evidence of Majorana zero modes – a manifestation of Majorana fermions and a crucial part of the proposed quantum system – was retracted. At the time of its original publication, it had been seen as important validation of the path Microsoft had taken.

Despite the very public setback, the team pushed forward. Microsoft had already abandoned the nanowires and semiconducting material behind the paper, so the team at Station Q chose to view the glass as half full: the paper’s retraction validated the choices Microsoft had made in the interim.

The materials stack the team ultimately designed and fabricated atom by atom combines indium arsenide, a semiconductor, and aluminum, a superconductor. When cooled to near absolute zero – to keep disruptive thermal vibrations and noise to a minimum – and tuned with magnetic fields, these devices would form topological superconducting nanowires with Majorana zero modes at the wires’ ends. Or so everyone hoped.

The breakthrough

Nayak remembers the breakthrough moment in the project. It was late at night in the U.S., just another regular workday for the European teams collecting data. The devices were tuned. The measurements were running. And then, there it was – a pattern in the oscillations that matched exactly what they’d been hoping to see. “It looked surprisingly good,” Nayak says.

They had just captured what they suspected could be evidence of a new phase of matter – a topological phase, defined by the emergence of Majorana zero modes. After that long-ago dinner of sparked inspiration and excited blackboard scribbling, followed by two decades of theorizing, revising, discarding entire fabrication methods, and building the team and process from the ground up, the team from Microsoft appeared to have taken a giant leap toward a new quantum era.

Nayak knew this data was too important to sit on. An impromptu meeting was called. It didn’t matter that it was 11 p.m. for the West Coast team. “Everyone needs to understand this,” Nayak remembers.

The team certainly had more work to do – replications, validations, independent checks. But for the first time, they could see it clearly: the qubit they’d envisioned more than 20 years earlier wasn’t purely theoretical anymore – they finally had compelling evidence it might be real.

The team invited an external council of quantum experts to review the results in detail and offer feedback and validation for the discovery. Then they got to work harnessing the properties of a topological qubit and combining it with control and measurement components to create what became the Majorana 1, a processor that can fit in the palm of your hand. It holds eight topological qubits and their surrounding control electronics, and is designed to scale to a million qubits.

The million number is crucial, as it’s a widely understood threshold for quantum computers to solve currently unsolvable problems. Every computer in the world working together could not accomplish what a computer with a million qubits will some day be able to do in minutes.

Qubits under control

Traditional approaches to controlling qubits have used analog wires – one for every qubit. That would get impossibly unwieldy when dealing with a million qubits. Microsoft instead designed a system that sends digital signals distributed as voltage pulses to the chips, controlling them at the physical level. Changes in the qubit are measured using microwaves.

The Majorana 1 also has to be cold enough for superconductivity, so it is housed inside a dilution refrigerator that can reach temperatures 100 times colder than those found in deep space. But cold is not enough; the device must also be immersed in a magnetic field a thousand times more powerful than the magnetic field of the Earth. The dilution refrigerator also houses amplifiers originally developed for radio astronomy but now used to listen to quantum devices.

And of course the chip cannot act alone. Even before there was proof of Majoranas, computer scientists like Microsoft Technical Fellow Krysta Svore were developing software that would someday run on the quantum computers that would harness them.

Svore’s fascination with quantum computing began in college, when she heard about it in a seminar and was drawn to the notion that “something we intuitively think is a hard problem can be solved by a different mode of computation.” But quantum was a new field with few opportunities when she began her career, so she focused on machine learning instead, joining Microsoft Research in 2007. She and her team developed the algorithm that initially powered Microsoft’s internet search engine Bing.

Then she met Freedman and got the opportunity she had always wanted. Svore was tapped to lead Microsoft Research’s Quantum Architectures and Computation Group, known as QuArC, which focused on building the systems that would run on quantum computers and could integrate them with AI and classical computers. In 2017, Svore and her colleagues introduced Microsoft’s Quantum Development Kit and Q#, a new programming language that would allow engineers to write their own quantum algorithms.

Svore likens Microsoft’s quantum computing journey to the familiar tale of the tortoise and the hare. Because Microsoft has been working in parallel on all the components needed to complete a full-stack system, progress will start to speed up from here, as the work continues with the creation of more complex devices and better qubits. “The tortoise ends up taking over, but not immediately,” she says. “Because it’s a harder path initially and becomes the easier path later. The other pieces of the full system, the full quantum computer, already exist. The hardest part was this new phase of matter and getting this qubit to prove out.”

Nayak reflects on the path so far with a sense of humility and awe – not only the physics, but at the team that made it possible. The 2005 version of the team – or even of himself – couldn’t have done it. The equipment didn’t exist. The required skills hadn’t yet been invented. The collaborative structure wasn’t there.

Over time, Microsoft’s quantum group has transformed from a scrappy band of researchers and physicists theorizing about what could perhaps be done into a multidisciplinary operation capable of solving one of the hardest engineering problems in human history. “The things we had to discover and invent to make this happen could have only happened through this journey,” he says.

“There’s no way three people writing a paper in 2004 could do this,” he continues. “Being able to stand there holding a chip took over 100 people. That team didn’t exist 10 or 15 years ago, but it was able to work together to accomplish something that no individual could have done.”