Microsoft, Broadcom and TransDigm were pitched on May 29, 2026, by Argent Capital Management portfolio manager Jed Ellerbroek as “Hot Picks” on BNN Bloomberg, with artificial intelligence demand, cloud capacity shortages, custom chip spending and aerospace aftermarket economics forming the core of the investment case. That may sound like another sunny tour through the AI trade, but the interview was more useful than the format suggests. It showed how the market’s AI story has moved from speculative wonder to a harder question: which companies can turn the infrastructure boom into durable earnings power?
The answer, if Ellerbroek is right, is not simply “whoever says AI the most.” It is companies with distribution, customer lock-in, scarce technical capacity, and pricing leverage in markets where demand is running ahead of supply. That distinction matters for Windows users and IT departments because the same forces lifting Microsoft’s stock are reshaping licensing, cloud architecture, software procurement and the cost of modern computing.
The first phase of the generative AI boom was indiscriminate. Investors bought the obvious hardware suppliers, then the cloud platforms, then anything with a credible story about automation, inference, data centers or productivity. That phase was intoxicating but sloppy, because it treated AI as a rising tide that would lift every enterprise software vendor, every chip designer and every infrastructure provider at roughly the same speed.
Ellerbroek’s comments point to a more mature market. He is still bullish on AI demand, but the logic is selective: Microsoft wins because it owns enterprise distribution and Azure capacity; Broadcom wins because custom accelerators are becoming strategic assets for hyperscalers; TransDigm wins for reasons that have almost nothing to do with AI, namely the stubbornly profitable physics of aerospace parts and aftermarket demand. The common thread is not artificial intelligence itself. It is leverage over scarce systems.
That is the piece investors and IT buyers both need to understand. AI is not just a product category; it is a capital allocation machine. It pushes Microsoft to build data centers, pushes Google and Meta toward custom silicon, pushes enterprises to renegotiate cloud commitments, and pushes software companies to prove that AI features are more than margin-diluting add-ons.
This is why the interview lands at an interesting moment. Microsoft’s latest reported Azure growth gives the bulls something concrete, while Broadcom’s upcoming earnings report gives the market another test of whether custom AI silicon is becoming a second center of gravity beside Nvidia. The old AI question was whether the technology was real. The new one is who gets paid after everyone agrees that it is.
But Microsoft is not an ordinary software company waiting to be attacked from the outside. It is the operating layer for much of corporate work: Windows endpoints, Microsoft 365 identities, Teams collaboration, SharePoint content, Power Platform automation, GitHub development and Azure infrastructure. That gives Microsoft a rare advantage in the AI transition. It does not have to persuade customers to move their data into a new universe; much of that data, identity and workflow gravity is already sitting inside Microsoft’s estate.
That is why the “death of software” narrative has always been too blunt. AI may compress the value of narrow workflow apps that perform a single repeatable task, especially if those tasks can be absorbed into broader platforms. But Microsoft’s business is not built on one task. It is built on being the sanctioned environment through which enterprises authenticate users, manage devices, store documents, collaborate, analyze data, and increasingly, deploy AI agents.
For WindowsForum readers, this is the practical implication: Microsoft’s AI strategy is inseparable from its platform strategy. Copilot is not merely a chatbot bolted onto Office. It is a licensing wedge, a cloud consumption driver, a data governance challenge and a reason for companies to revisit whether their endpoint, identity and productivity stacks are modern enough to support AI at scale.
That does not make Microsoft invulnerable. It does mean that the company’s risk profile is different from a software vendor trying to defend one workflow against a clever AI-native startup. Microsoft can lose some product-level battles and still win the platform-level war if AI adoption makes Azure, Entra, Microsoft 365 and GitHub more central to enterprise architecture.
But Azure’s growth should not be read as a simple victory lap. AI growth is expensive growth. Microsoft’s cloud acceleration comes with enormous capital expenditures, long-term data center commitments, power constraints, GPU and accelerator procurement, networking demands, and pressure to make AI workloads profitable at scale. The company is not just selling higher-margin software subscriptions; it is building the industrial plant required to run the next generation of computing.
That is the tension at the center of Microsoft’s current valuation. Bulls see a company converting its enterprise relationships into cloud and AI revenue. Skeptics see an arms race in which Microsoft must spend heavily just to stay in contention against Amazon, Google, Meta and specialized AI infrastructure players. Both views can be true at the same time.
For enterprise customers, the capacity story is more than Wall Street theater. If demand for AI compute continues to exceed supply, cloud purchasing becomes less like buying commodity storage and more like securing strategic infrastructure. Reserved capacity, committed spend, regional availability, data residency, GPU access and model hosting choices all become board-level IT concerns.
That is why new entrants into cloud computing do not automatically weaken Microsoft’s position. Ellerbroek brushed off the idea of Meta entering the cloud wars by describing the market as a “big ocean.” In the near term, that is plausible. When demand exceeds supply, additional capacity can validate the market rather than destroy pricing.
The longer-term question is different. If hyperscalers and AI labs build more of their own infrastructure, and if specialized neo-clouds offer cheaper or more flexible AI capacity, Microsoft will have to prove that Azure is not just available, but operationally superior. Its advantage will not be raw compute alone. It will be the combination of compute, enterprise controls, compliance, identity, developer tooling and procurement familiarity.
That is exactly the distinction IT departments should be making now. The question is not whether a vendor has AI features. It is whether those features improve the underlying product enough to justify higher cost, deeper integration or longer contractual commitment. A generated summary is not a moat. A workflow that saves hours, respects permissions, preserves auditability and fits into existing governance may be.
Microsoft sits on the favorable side of that line because it can embed AI into work patterns that already exist. An AI assistant inside Outlook, Excel, Teams, Windows or Visual Studio has a distribution advantage that an independent assistant does not. The friction is lower, the procurement path is clearer, and the trust model starts from an existing relationship.
But that same advantage creates a trap. Microsoft can easily ship AI everywhere; the harder task is making AI feel necessary rather than noisy. Many Windows users have already seen the awkward early phase of Copilot branding, where AI surfaced across products faster than organizations could decide what they wanted it to do. The market may reward AI ubiquity, but customers reward usefulness.
That gap between feature availability and operational value is where CIOs will spend the next two years. They will ask whether Copilot improves document work, whether GitHub Copilot materially accelerates development, whether AI agents can be governed, whether sensitive data leaks into prompts, and whether the productivity gains offset licensing and infrastructure costs. The winners in software will be the companies that can answer those questions with measurable outcomes rather than demo-stage enthusiasm.
This is why “AI disruption” is both a threat and a subsidy for incumbents. It threatens software vendors whose products are thin wrappers around information retrieval or repetitive task execution. It subsidizes vendors that already own the system of record, the collaboration layer, or the security boundary. Microsoft owns several of those layers, which is why the market keeps returning to it even after bouts of AI skepticism.
Nvidia sells a general-purpose accelerated computing platform with GPUs, networking, software libraries and a developer ecosystem that has become the default for frontier AI. Broadcom, by contrast, benefits when the largest hyperscalers decide that some workloads are important enough, predictable enough and expensive enough to justify custom silicon. That is not a rejection of Nvidia. It is what happens when AI spending gets so large that optimization becomes unavoidable.
Custom chips are not new, but the economics have changed. When a company is spending billions on AI infrastructure, even modest improvements in performance per watt, networking efficiency or workload-specific throughput can become strategically meaningful. Google’s TPUs were once a specialized internal advantage. In the current market, TPU capacity has become part of a broader cloud and AI supply story.
Broadcom’s appeal is that it sells picks and shovels to the companies wealthy enough to design their own mines. Its custom silicon business does not depend on persuading every enterprise to become a chip designer. It depends on a small number of hyperscale customers placing enormous, long-duration bets on AI infrastructure.
That concentration cuts both ways. A few giant customers can create extraordinary revenue visibility, but they can also create dependency. If one hyperscaler changes architecture, delays deployment, insources more design work or shifts capital plans, the impact can be large. Broadcom’s upcoming earnings guidance therefore matters not just for the stock, but for what it reveals about the durability of custom accelerator demand.
Ellerbroek’s emphasis on next-quarter guidance is exactly right. In this market, the past quarter is less interesting than the order book. Investors want to know whether AI infrastructure demand is still pulling forward, whether TPU-related deployments are expanding, and whether Broadcom’s software business remains a stabilizing second engine rather than a distraction.
Ellerbroek’s “more demand than supply” line is the central investment claim. If true, it explains why large cloud providers remain strong despite new entrants, why AI chip demand has stayed resilient, and why software investors are willing to revisit names they had recently punished. Scarcity changes behavior. Customers commit earlier, providers spend more aggressively, and suppliers gain pricing power.
But scarcity also hides mistakes. When everyone is capacity constrained, nearly every supplier looks smart. The real test comes when supply catches up, when depreciation hits, when models become more efficient, or when enterprises decide that not every workflow needs expensive AI inference. The AI boom may be real and still produce overbuilt pockets.
For Windows and enterprise IT professionals, that means cloud strategy cannot be based on hype cycles alone. A company betting heavily on Azure AI services, Google Cloud TPUs, AWS Bedrock, OpenAI APIs or private GPU clusters is also making assumptions about latency, data gravity, vendor lock-in, regulatory exposure and future pricing. The cheapest pilot can become the most expensive architecture if it traps data or workflows in the wrong place.
Microsoft has an advantage here because enterprise buyers often prefer an integrated story. They can buy productivity software, identity, endpoint management, developer tools, security products and cloud services from one vendor. That simplicity has procurement value, especially for large organizations tired of stitching together AI governance across a dozen platforms.
Yet the same bundling that helps Microsoft can frustrate customers. If AI becomes another reason to accept higher Microsoft 365 tiers, more Azure commitments and deeper dependence on Redmond’s roadmap, some IT leaders will look for counterweights. Open-source models, multi-cloud inference, specialized AI clouds and custom deployments will all get consideration, not necessarily because they are simpler, but because they preserve bargaining power.
Ellerbroek pointed to growth across defense, commercial aerospace, new equipment and aftermarket sales. He also emphasized acquisitions, including companies tied to aircraft parts and repairs. This is classic TransDigm logic: buy specialized aerospace component businesses, focus on proprietary products, and benefit from pricing power where certification barriers and safety requirements limit competition.
That model has attracted criticism over the years, particularly around pricing in defense supply chains. But as an investment case, it explains why aerospace aftermarket businesses command attention. Aircraft remain in service for long periods, fleets require maintenance regardless of macro sentiment, and replacement parts are not usually bought through the same competitive dynamics as consumer electronics.
The comparison with Microsoft and Broadcom is useful because all three picks are ultimately about control points. Microsoft controls enterprise software and cloud distribution. Broadcom controls specialized silicon and networking relationships inside hyperscale AI systems. TransDigm controls narrow but valuable aerospace component niches where customers often have few alternatives.
This is the broader lesson behind the segment. Markets reward growth, but they reward defensible growth more. AI demand is spectacular, but without a moat it can become a spending treadmill. Aerospace aftermarket growth is less glamorous, but its durability comes from regulation, certification and installed base.
For IT readers, TransDigm is also a reminder not to confuse “technology story” with “software story.” Modern aerospace is a deeply technical supply chain, just as AI is an industrial supply chain. The companies that win are not always the ones closest to the user interface. Often they are the ones buried deep enough in the system that customers cannot easily replace them.
That divergence is visible in Microsoft’s position. Investors may celebrate Azure growth and AI revenue run rates, but CIOs see capacity planning, licensing complexity and risk management. Copilot can be a productivity tool, but it can also expose poor data hygiene. Azure AI can accelerate development, but it can also deepen reliance on a cloud provider whose pricing and roadmap customers do not control.
Broadcom creates a different kind of divergence. Investors see custom silicon demand as evidence that AI infrastructure spending is broadening beyond Nvidia. Enterprises may see that same trend and wonder whether the cloud market is fragmenting into specialized hardware backends with different performance profiles, pricing models and availability constraints. The abstraction layer is supposed to hide that complexity, but at AI scale, hardware choices increasingly leak into software architecture.
The practical result is that AI procurement is becoming more technical, not less. Buying “AI capability” is meaningless unless a company understands where data sits, which models are used, what compute is required, how access is controlled, and whether the vendor can meet performance and compliance requirements. The board may want an AI strategy. The IT department has to turn that phrase into architecture.
This is where Microsoft’s Windows ecosystem still matters. Endpoints remain the place where users encounter AI, data is created, identities are enforced and policies succeed or fail. The cloud may run the models, but the desktop is where productivity claims become visible. If Microsoft can make Windows, Microsoft 365, Edge, Defender, Intune and Copilot feel like one governed AI workspace, it strengthens the case for staying inside its stack.
If it cannot, AI becomes another layer of fragmentation. Users will bring their own tools, departments will buy their own assistants, developers will wire up their own APIs, and security teams will spend the next several years cleaning up the mess. That is the administrative reality behind the market euphoria.
Microsoft will have to show that AI spending produces durable cloud growth and higher-value software relationships, not just ballooning capital expenditure. Broadcom will have to show that custom silicon demand is broadening, repeatable and profitable. Software vendors will have to show that AI features increase customer value rather than merely defending existing contracts.
This is healthy. The first wave of AI excitement needed imagination because the technology’s possibilities were genuinely new to most users. The second wave needs discipline because the bills are now enormous. Data centers are not PowerPoint slides. GPUs, TPUs, networking gear and electricity costs must eventually be matched by revenue that customers are willing to keep paying.
For WindowsForum’s audience, the same discipline applies at organizational scale. Pilots are easy. Production is hard. A company can let a few teams experiment with AI tools and declare momentum, but enterprise deployment requires identity design, retention policies, prompt logging, eDiscovery planning, endpoint controls, data classification and user training.
The vendor that makes those boring requirements easier has a better chance of turning AI hype into platform loyalty. That is Microsoft’s opportunity. It is also Microsoft’s burden, because no company has more enterprise surface area through which AI confusion can spread.
Microsoft has enterprise trust, distribution and Azure scale. Broadcom has custom silicon relationships and deep infrastructure exposure. TransDigm has certified aerospace parts and aftermarket niches. Each company is positioned around constraints that customers cannot solve overnight.
That is why the “AI demand” headline is only partly adequate. AI is the accelerant, but scarcity is the investment thesis. Cloud capacity is scarce. Custom accelerator expertise is scarce. Enterprise-grade AI governance is scarce. Certified aerospace components are scarce. Investors are following the bottlenecks.
The risk is that bottlenecks move. AI models may become more efficient. Hyperscalers may rebalance between Nvidia GPUs and custom chips. Enterprises may slow deployment if returns disappoint. Regulators may complicate data usage. Cloud providers may discover that customers want AI features but resist the full cost.
Still, the current evidence supports a market that is becoming more selective rather than less enthusiastic. The AI boom has not ended; it has become more demanding. That is usually what happens when a technology shifts from story to infrastructure.
The answer, if Ellerbroek is right, is not simply “whoever says AI the most.” It is companies with distribution, customer lock-in, scarce technical capacity, and pricing leverage in markets where demand is running ahead of supply. That distinction matters for Windows users and IT departments because the same forces lifting Microsoft’s stock are reshaping licensing, cloud architecture, software procurement and the cost of modern computing.
The AI Trade Has Stopped Being One Trade
The first phase of the generative AI boom was indiscriminate. Investors bought the obvious hardware suppliers, then the cloud platforms, then anything with a credible story about automation, inference, data centers or productivity. That phase was intoxicating but sloppy, because it treated AI as a rising tide that would lift every enterprise software vendor, every chip designer and every infrastructure provider at roughly the same speed.Ellerbroek’s comments point to a more mature market. He is still bullish on AI demand, but the logic is selective: Microsoft wins because it owns enterprise distribution and Azure capacity; Broadcom wins because custom accelerators are becoming strategic assets for hyperscalers; TransDigm wins for reasons that have almost nothing to do with AI, namely the stubbornly profitable physics of aerospace parts and aftermarket demand. The common thread is not artificial intelligence itself. It is leverage over scarce systems.
That is the piece investors and IT buyers both need to understand. AI is not just a product category; it is a capital allocation machine. It pushes Microsoft to build data centers, pushes Google and Meta toward custom silicon, pushes enterprises to renegotiate cloud commitments, and pushes software companies to prove that AI features are more than margin-diluting add-ons.
This is why the interview lands at an interesting moment. Microsoft’s latest reported Azure growth gives the bulls something concrete, while Broadcom’s upcoming earnings report gives the market another test of whether custom AI silicon is becoming a second center of gravity beside Nvidia. The old AI question was whether the technology was real. The new one is who gets paid after everyone agrees that it is.
Microsoft Is No Longer Selling Software So Much as Permission to Modernize
Ellerbroek’s Microsoft case begins with a familiar claim: the stock had been “beaten down” by fears that AI would disrupt software. That fear is not imaginary. If generative AI can draft emails, summarize files, write code, query databases, prepare slides and automate workflows, then a lot of traditional software seats start looking less sacred than they did five years ago.But Microsoft is not an ordinary software company waiting to be attacked from the outside. It is the operating layer for much of corporate work: Windows endpoints, Microsoft 365 identities, Teams collaboration, SharePoint content, Power Platform automation, GitHub development and Azure infrastructure. That gives Microsoft a rare advantage in the AI transition. It does not have to persuade customers to move their data into a new universe; much of that data, identity and workflow gravity is already sitting inside Microsoft’s estate.
That is why the “death of software” narrative has always been too blunt. AI may compress the value of narrow workflow apps that perform a single repeatable task, especially if those tasks can be absorbed into broader platforms. But Microsoft’s business is not built on one task. It is built on being the sanctioned environment through which enterprises authenticate users, manage devices, store documents, collaborate, analyze data, and increasingly, deploy AI agents.
For WindowsForum readers, this is the practical implication: Microsoft’s AI strategy is inseparable from its platform strategy. Copilot is not merely a chatbot bolted onto Office. It is a licensing wedge, a cloud consumption driver, a data governance challenge and a reason for companies to revisit whether their endpoint, identity and productivity stacks are modern enough to support AI at scale.
That does not make Microsoft invulnerable. It does mean that the company’s risk profile is different from a software vendor trying to defend one workflow against a clever AI-native startup. Microsoft can lose some product-level battles and still win the platform-level war if AI adoption makes Azure, Entra, Microsoft 365 and GitHub more central to enterprise architecture.
Azure’s 40 Percent Growth Is a Capacity Story Disguised as a Revenue Line
Ellerbroek cited Azure as growing around 40 percent, which lines up with Microsoft’s latest fiscal third-quarter reporting. The number matters because it suggests that the cloud business is not merely riding a temporary sentiment wave. It is absorbing a structural shift in compute demand, and that demand is now large enough to show up clearly in Microsoft’s financials.But Azure’s growth should not be read as a simple victory lap. AI growth is expensive growth. Microsoft’s cloud acceleration comes with enormous capital expenditures, long-term data center commitments, power constraints, GPU and accelerator procurement, networking demands, and pressure to make AI workloads profitable at scale. The company is not just selling higher-margin software subscriptions; it is building the industrial plant required to run the next generation of computing.
That is the tension at the center of Microsoft’s current valuation. Bulls see a company converting its enterprise relationships into cloud and AI revenue. Skeptics see an arms race in which Microsoft must spend heavily just to stay in contention against Amazon, Google, Meta and specialized AI infrastructure players. Both views can be true at the same time.
For enterprise customers, the capacity story is more than Wall Street theater. If demand for AI compute continues to exceed supply, cloud purchasing becomes less like buying commodity storage and more like securing strategic infrastructure. Reserved capacity, committed spend, regional availability, data residency, GPU access and model hosting choices all become board-level IT concerns.
That is why new entrants into cloud computing do not automatically weaken Microsoft’s position. Ellerbroek brushed off the idea of Meta entering the cloud wars by describing the market as a “big ocean.” In the near term, that is plausible. When demand exceeds supply, additional capacity can validate the market rather than destroy pricing.
The longer-term question is different. If hyperscalers and AI labs build more of their own infrastructure, and if specialized neo-clouds offer cheaper or more flexible AI capacity, Microsoft will have to prove that Azure is not just available, but operationally superior. Its advantage will not be raw compute alone. It will be the combination of compute, enterprise controls, compliance, identity, developer tooling and procurement familiarity.
Software Is Not Dead, but the Lazy Seat Is in Trouble
The most useful part of Ellerbroek’s software commentary was its refusal to make a sweeping call. Some software companies will struggle because AI will replace or diminish their usefulness. Others will become more valuable because AI makes their products better, stickier and more central to customer workflows.That is exactly the distinction IT departments should be making now. The question is not whether a vendor has AI features. It is whether those features improve the underlying product enough to justify higher cost, deeper integration or longer contractual commitment. A generated summary is not a moat. A workflow that saves hours, respects permissions, preserves auditability and fits into existing governance may be.
Microsoft sits on the favorable side of that line because it can embed AI into work patterns that already exist. An AI assistant inside Outlook, Excel, Teams, Windows or Visual Studio has a distribution advantage that an independent assistant does not. The friction is lower, the procurement path is clearer, and the trust model starts from an existing relationship.
But that same advantage creates a trap. Microsoft can easily ship AI everywhere; the harder task is making AI feel necessary rather than noisy. Many Windows users have already seen the awkward early phase of Copilot branding, where AI surfaced across products faster than organizations could decide what they wanted it to do. The market may reward AI ubiquity, but customers reward usefulness.
That gap between feature availability and operational value is where CIOs will spend the next two years. They will ask whether Copilot improves document work, whether GitHub Copilot materially accelerates development, whether AI agents can be governed, whether sensitive data leaks into prompts, and whether the productivity gains offset licensing and infrastructure costs. The winners in software will be the companies that can answer those questions with measurable outcomes rather than demo-stage enthusiasm.
This is why “AI disruption” is both a threat and a subsidy for incumbents. It threatens software vendors whose products are thin wrappers around information retrieval or repetitive task execution. It subsidizes vendors that already own the system of record, the collaboration layer, or the security boundary. Microsoft owns several of those layers, which is why the market keeps returning to it even after bouts of AI skepticism.
Broadcom Is the AI Boom’s Less Glamorous Power Broker
Nvidia remains the defining company of the AI hardware era, but Broadcom’s role is increasingly difficult to ignore. Ellerbroek’s Broadcom pick rests on custom accelerators, especially the company’s work around Google’s TPU ecosystem and reported or discussed relationships involving other major AI players. That puts Broadcom in a different lane from the Nvidia trade.Nvidia sells a general-purpose accelerated computing platform with GPUs, networking, software libraries and a developer ecosystem that has become the default for frontier AI. Broadcom, by contrast, benefits when the largest hyperscalers decide that some workloads are important enough, predictable enough and expensive enough to justify custom silicon. That is not a rejection of Nvidia. It is what happens when AI spending gets so large that optimization becomes unavoidable.
Custom chips are not new, but the economics have changed. When a company is spending billions on AI infrastructure, even modest improvements in performance per watt, networking efficiency or workload-specific throughput can become strategically meaningful. Google’s TPUs were once a specialized internal advantage. In the current market, TPU capacity has become part of a broader cloud and AI supply story.
Broadcom’s appeal is that it sells picks and shovels to the companies wealthy enough to design their own mines. Its custom silicon business does not depend on persuading every enterprise to become a chip designer. It depends on a small number of hyperscale customers placing enormous, long-duration bets on AI infrastructure.
That concentration cuts both ways. A few giant customers can create extraordinary revenue visibility, but they can also create dependency. If one hyperscaler changes architecture, delays deployment, insources more design work or shifts capital plans, the impact can be large. Broadcom’s upcoming earnings guidance therefore matters not just for the stock, but for what it reveals about the durability of custom accelerator demand.
Ellerbroek’s emphasis on next-quarter guidance is exactly right. In this market, the past quarter is less interesting than the order book. Investors want to know whether AI infrastructure demand is still pulling forward, whether TPU-related deployments are expanding, and whether Broadcom’s software business remains a stabilizing second engine rather than a distraction.
The Cloud Wars Are Becoming a Supply Chain War
The old cloud wars were about regions, services, developer mindshare and enterprise migration. The new cloud wars are about power, accelerators, networking, cooling, land, supply agreements and the ability to turn capital expenditure into usable capacity. That shift is why Microsoft, Broadcom and Nvidia can all be beneficiaries of the same demand curve without occupying the same part of the stack.Ellerbroek’s “more demand than supply” line is the central investment claim. If true, it explains why large cloud providers remain strong despite new entrants, why AI chip demand has stayed resilient, and why software investors are willing to revisit names they had recently punished. Scarcity changes behavior. Customers commit earlier, providers spend more aggressively, and suppliers gain pricing power.
But scarcity also hides mistakes. When everyone is capacity constrained, nearly every supplier looks smart. The real test comes when supply catches up, when depreciation hits, when models become more efficient, or when enterprises decide that not every workflow needs expensive AI inference. The AI boom may be real and still produce overbuilt pockets.
For Windows and enterprise IT professionals, that means cloud strategy cannot be based on hype cycles alone. A company betting heavily on Azure AI services, Google Cloud TPUs, AWS Bedrock, OpenAI APIs or private GPU clusters is also making assumptions about latency, data gravity, vendor lock-in, regulatory exposure and future pricing. The cheapest pilot can become the most expensive architecture if it traps data or workflows in the wrong place.
Microsoft has an advantage here because enterprise buyers often prefer an integrated story. They can buy productivity software, identity, endpoint management, developer tools, security products and cloud services from one vendor. That simplicity has procurement value, especially for large organizations tired of stitching together AI governance across a dozen platforms.
Yet the same bundling that helps Microsoft can frustrate customers. If AI becomes another reason to accept higher Microsoft 365 tiers, more Azure commitments and deeper dependence on Redmond’s roadmap, some IT leaders will look for counterweights. Open-source models, multi-cloud inference, specialized AI clouds and custom deployments will all get consideration, not necessarily because they are simpler, but because they preserve bargaining power.
TransDigm Is the Reminder That Not Every Durable Trend Has a Chatbot
TransDigm might seem like the odd name in a tech-heavy “Hot Picks” segment, but its inclusion is instructive. The company’s aerospace and defense business benefits from a different kind of scarcity: certified parts, long aircraft lifecycles, limited suppliers and aftermarket demand that can persist for decades. In a market obsessed with AI, that is almost refreshingly old-fashioned.Ellerbroek pointed to growth across defense, commercial aerospace, new equipment and aftermarket sales. He also emphasized acquisitions, including companies tied to aircraft parts and repairs. This is classic TransDigm logic: buy specialized aerospace component businesses, focus on proprietary products, and benefit from pricing power where certification barriers and safety requirements limit competition.
That model has attracted criticism over the years, particularly around pricing in defense supply chains. But as an investment case, it explains why aerospace aftermarket businesses command attention. Aircraft remain in service for long periods, fleets require maintenance regardless of macro sentiment, and replacement parts are not usually bought through the same competitive dynamics as consumer electronics.
The comparison with Microsoft and Broadcom is useful because all three picks are ultimately about control points. Microsoft controls enterprise software and cloud distribution. Broadcom controls specialized silicon and networking relationships inside hyperscale AI systems. TransDigm controls narrow but valuable aerospace component niches where customers often have few alternatives.
This is the broader lesson behind the segment. Markets reward growth, but they reward defensible growth more. AI demand is spectacular, but without a moat it can become a spending treadmill. Aerospace aftermarket growth is less glamorous, but its durability comes from regulation, certification and installed base.
For IT readers, TransDigm is also a reminder not to confuse “technology story” with “software story.” Modern aerospace is a deeply technical supply chain, just as AI is an industrial supply chain. The companies that win are not always the ones closest to the user interface. Often they are the ones buried deep enough in the system that customers cannot easily replace them.
The Investor Case and the IT Case Are Starting to Diverge
Wall Street likes the AI story because it promises growth across cloud, semiconductors and software. IT departments have a more complicated relationship with the same trend. They have to fund it, secure it, govern it and explain why it is worth the disruption.That divergence is visible in Microsoft’s position. Investors may celebrate Azure growth and AI revenue run rates, but CIOs see capacity planning, licensing complexity and risk management. Copilot can be a productivity tool, but it can also expose poor data hygiene. Azure AI can accelerate development, but it can also deepen reliance on a cloud provider whose pricing and roadmap customers do not control.
Broadcom creates a different kind of divergence. Investors see custom silicon demand as evidence that AI infrastructure spending is broadening beyond Nvidia. Enterprises may see that same trend and wonder whether the cloud market is fragmenting into specialized hardware backends with different performance profiles, pricing models and availability constraints. The abstraction layer is supposed to hide that complexity, but at AI scale, hardware choices increasingly leak into software architecture.
The practical result is that AI procurement is becoming more technical, not less. Buying “AI capability” is meaningless unless a company understands where data sits, which models are used, what compute is required, how access is controlled, and whether the vendor can meet performance and compliance requirements. The board may want an AI strategy. The IT department has to turn that phrase into architecture.
This is where Microsoft’s Windows ecosystem still matters. Endpoints remain the place where users encounter AI, data is created, identities are enforced and policies succeed or fail. The cloud may run the models, but the desktop is where productivity claims become visible. If Microsoft can make Windows, Microsoft 365, Edge, Defender, Intune and Copilot feel like one governed AI workspace, it strengthens the case for staying inside its stack.
If it cannot, AI becomes another layer of fragmentation. Users will bring their own tools, departments will buy their own assistants, developers will wire up their own APIs, and security teams will spend the next several years cleaning up the mess. That is the administrative reality behind the market euphoria.
The AI Winners Will Be Judged by Boring Metrics
The most important AI metrics over the next year may not be the flashiest ones. Model benchmarks, demo videos and keynote promises will matter less than utilization, gross margin, backlog conversion, renewal rates and customer retention. The market is gradually moving from imagination to accounting.Microsoft will have to show that AI spending produces durable cloud growth and higher-value software relationships, not just ballooning capital expenditure. Broadcom will have to show that custom silicon demand is broadening, repeatable and profitable. Software vendors will have to show that AI features increase customer value rather than merely defending existing contracts.
This is healthy. The first wave of AI excitement needed imagination because the technology’s possibilities were genuinely new to most users. The second wave needs discipline because the bills are now enormous. Data centers are not PowerPoint slides. GPUs, TPUs, networking gear and electricity costs must eventually be matched by revenue that customers are willing to keep paying.
For WindowsForum’s audience, the same discipline applies at organizational scale. Pilots are easy. Production is hard. A company can let a few teams experiment with AI tools and declare momentum, but enterprise deployment requires identity design, retention policies, prompt logging, eDiscovery planning, endpoint controls, data classification and user training.
The vendor that makes those boring requirements easier has a better chance of turning AI hype into platform loyalty. That is Microsoft’s opportunity. It is also Microsoft’s burden, because no company has more enterprise surface area through which AI confusion can spread.
The Real Signal in Ellerbroek’s Picks Is Scarcity
Ellerbroek’s three picks look different on the surface, but they tell one story about where markets believe pricing power will sit in 2026. The winners are not simply the companies with exposure to fashionable demand. They are companies that can meet demand through assets competitors cannot quickly reproduce.Microsoft has enterprise trust, distribution and Azure scale. Broadcom has custom silicon relationships and deep infrastructure exposure. TransDigm has certified aerospace parts and aftermarket niches. Each company is positioned around constraints that customers cannot solve overnight.
That is why the “AI demand” headline is only partly adequate. AI is the accelerant, but scarcity is the investment thesis. Cloud capacity is scarce. Custom accelerator expertise is scarce. Enterprise-grade AI governance is scarce. Certified aerospace components are scarce. Investors are following the bottlenecks.
The risk is that bottlenecks move. AI models may become more efficient. Hyperscalers may rebalance between Nvidia GPUs and custom chips. Enterprises may slow deployment if returns disappoint. Regulators may complicate data usage. Cloud providers may discover that customers want AI features but resist the full cost.
Still, the current evidence supports a market that is becoming more selective rather than less enthusiastic. The AI boom has not ended; it has become more demanding. That is usually what happens when a technology shifts from story to infrastructure.
The Hot Picks Hide a Cooler Message for Buyers
Ellerbroek’s interview is most useful when read less as stock promotion and more as a map of pressure points that IT leaders will face over the next year. The names are public equities, but the forces behind them will show up in budgets, contracts, architectures and support tickets.- Microsoft’s AI advantage depends on whether enterprises treat Copilot, Azure and Microsoft 365 as a governed platform rather than a bundle of loosely connected features.
- Broadcom’s rise shows that hyperscalers are serious about custom accelerators, which could make AI cloud performance and pricing more dependent on hardware choices than traditional cloud abstractions suggest.
- Cloud demand exceeding supply gives providers leverage now, but it also raises the risk that customers lock in expensive commitments before their AI workloads are fully understood.
- Software vendors will increasingly be judged by whether AI features create measurable workflow value, not by whether they can add a chatbot to an existing interface.
- TransDigm’s inclusion is a reminder that durable pricing power often comes from overlooked infrastructure, certification and aftermarket control rather than headline-grabbing innovation.
- The practical AI winners for businesses will be the vendors that make governance, security, cost control and integration less painful, not merely the vendors with the most ambitious demos.
References
- Primary source: BNN Bloomberg
Published: 2026-05-29T20:30:37.764763
Hot Picks: Microsoft, Broadcom benefit from accelerating AI demand
AI demand is boosting software, cloud and semiconductor stocks as investors look for technology leaders with durable growth prospects.www.bnnbloomberg.ca
- Official source: microsoft.com
FY26 Q3 - Intelligent Cloud Performance - Investor Relations - Microsoft
FY26 Q3 - Intelligent Cloud Performance - Investor Relations - Microsoftwww.microsoft.com - Related coverage: investors.broadcom.com
Broadcom Inc. to Announce Second Quarter Fiscal Year 2026 Financial Results on Wednesday, June 3, 2026 | Broadcom Inc.
The Investor Relations website contains information about Broadcom Inc. 's business for stockholders, potential investors, and financial analysts.
investors.broadcom.com
- Related coverage: autais.com
Microsoft Q3 FY2026 決算、AI 事業ARR $37B（前年比+123%）・Azure 成長40% | AI・DXニュース・トピックス | Autais
Microsoft が 2026年4月29日に発表した FY26 第3四半期決算で、売上 $82.9B（前年比+18%）、Azure 成長率 40%、AI 事業 ARR が $37B（前年比+123%）に到達。Microsoft 365 Copilot 有料シートは 20M 超、新規シート追加は前年比+250%。年間設備投資は $190B 規模に拡大見通し。autais.com - Related coverage: marketbeat.com
AVGO Q2 2026 Earnings Report on 6/3/2026
Broadcom announced their Q2 2026 earnings on 6/3/2026. View AVGO's earnings results at MarketBeat.www.marketbeat.com
- Related coverage: tomshardware.com
Broadcom to supply Anthropic with 3.5 gigawatts of Google TPU capacity from 2027 — Claude pioneer says its annual revenue run rate has passed $30 billion
Securities filing confirms multi-year supply agreement runs through 2031.www.tomshardware.com
- Related coverage: en.bulios.com
Microsoft | Q3 2026: Revenue grows 18% and Azure accelerates 40%
Microsoft delivered another very strong set of numbers for the third fiscal quarter of 2026, confirming that cloud and AI are the main drivers of growth. Revenues grew 18% to $82.9 billion (15% at…en.bulios.com
- Related coverage: windowscentral.com
Microsoft is spending like a company in control, but the numbers tell a more complicated story
Microsoft’s growth keeps rising, but investors want to see whether massive AI spending can turn into real profit.
www.windowscentral.com
- Related coverage: savest-financial.com
