Introduction: The End of a Landmark AI Partnership
The AI industry was built on ambitious partnerships, but few were as closely watched as the multi-billion dollar agreement between Microsoft and the specialized hardware startup Nscale. In a move that sent shockwaves through the technology sector, Microsoft formally announced the termination of this landmark deal. This decision effectively halts the planned integration of Nscale's bespoke AI accelerators into the Azure cloud infrastructure, a project once hailed as the next frontier in hyperscale computing. This development is a major component of the broader strategic shift detailed in our parent analysis, Microsoft Exits Nscale AI Project, and it signals a profound realignment in how major cloud providers approach their hardware stack.
The cancellation is far more than a simple contract dispute; it is a bellwether for the entire AI hardware ecosystem. It raises critical questions about the viability of third-party silicon vendors in a world increasingly dominated by the in-house designs of tech giants like Microsoft, Google, and Amazon. For Nscale, the termination is an existential threat. For Microsoft, it represents a calculated, albeit costly, pivot towards vertical integration and self-reliance in the critical path of AI development. This article will provide a comprehensive deep dive into the cancellation, exploring the original vision of the partnership, the warning signs that preceded the collapse, the immediate fallout, and the long-term strategic implications for the cloud computing and AI industries as we look toward 2026 and beyond. We will deconstruct the corporate messaging, analyze the underlying technical and business drivers, and offer insights into what this means for enterprise customers, investors, and technology professionals navigating this volatile landscape.
The Original Vision: Unpacking the Nscale-Microsoft Partnership Goals
To understand the gravity of the cancellation, one must first appreciate the ambition of the original agreement. The partnership, detailed in its early stages as the Microsoft Nscale data centre deal, was not merely a procurement contract; it was a deep, co-engineering effort designed to solve one of the most pressing challenges in artificial intelligence: computational scaling. As large language models (LLMs) and other foundation models grew exponentially in size, the demand for processing power and memory bandwidth began to outstrip the capabilities of even the most advanced general-purpose GPUs. Nscale emerged as a promising solution to this bottleneck.
The startup's core innovation was a novel chip architecture, the 'Tensor Fabric Unit' (TFU), designed from the ground up for massive-scale neural network training and inference. Unlike traditional GPUs, which must balance graphics rendering with computation, Nscale's TFUs were pure AI accelerators. They featured a unique memory-on-chip design that minimized data movement, a major source of latency and power consumption. Furthermore, Nscale developed a proprietary high-speed interconnect, dubbed 'Synapse Link,' which allowed tens of thousands of TFUs to be linked together into a cohesive computational fabric, appearing to software as a single, colossal processing unit. This was the key differentiator: it promised to simplify the notoriously complex challenge of distributed training, potentially reducing training times for next-generation models from months to weeks.
Microsoft's vision was to integrate this technology directly into Azure's core infrastructure. The plan involved building specialized data center halls filled with Nscale hardware, offered to Azure customers as a premium 'AI Supercomputing' tier. This would have given Microsoft a powerful competitive advantage over rivals like AWS and Google Cloud, who were primarily reliant on off-the-shelf Nvidia GPUs or their own, more generalized, custom silicon. The partnership was a classic 'make versus buy' decision where Microsoft initially chose 'buy' to accelerate its time-to-market. The deal encompassed a multi-year, multi-billion dollar commitment for Nscale chips, joint R&D on next-generation architectures, and deep software integration between Nscale's compiler stack and Microsoft's Azure AI platform. It was a symbiotic relationship: Nscale secured a whale of a customer to fund its roadmap, and Microsoft gained a bleeding-edge hardware advantage to power its own services like Copilot and its enterprise AI offerings.
Cracks in the Foundation: Early Warning Signs and Strategic Misalignment
The dissolution of a partnership this significant rarely happens overnight. With the benefit of hindsight, a series of warning signs pointed to growing friction and a divergence of strategic interests long before the official announcement. These issues spanned technical execution, financial pressures, and a fundamental shift in Microsoft's own internal hardware ambitions. The first public indicators were subtle but significant: missed timelines. The initial pilot deployment of Nscale's first-generation hardware in select Azure data centers was quietly delayed by two quarters. While both companies attributed this to 'supply chain optimizations,' industry insiders whispered of yield issues with Nscale's advanced packaging process and lower-than-expected performance on key benchmarks.
A more telling sign was the departure of several key engineering leads from Nscale who had been instrumental in designing the TFU architecture. Such departures are common in the high-pressure startup world, but the loss of a core architecture team during a critical customer integration phase suggested deeper problems. It hinted at potential disagreements over the technical roadmap or frustration with the pace of progress. Simultaneously, Microsoft's own rhetoric began to change. In public presentations and investor calls, executives began to speak more forcefully about their long-term investments in custom silicon, particularly the 'Maia' AI accelerator and 'Athena' networking chip. This was a clear signal that while the Nscale partnership was the immediate plan, Microsoft was aggressively developing a Plan B, hedging its bets on its own in-house capabilities.
Financial pressures also mounted. Nscale, despite its massive Microsoft contract, was burning through cash at an alarming rate to fund its R&D and manufacturing ramp-up. The delayed deployment meant delayed revenue, creating a potential cash flow crunch. This may have led Nscale to push for revised contract terms or faster payments, creating friction with Microsoft's procurement and finance departments. The final and most crucial factor was strategic misalignment. The AI landscape was evolving faster than anyone anticipated. When the deal was signed, the primary challenge was massive-scale training. However, the rise of generative AI shifted the focus significantly towards cost-effective, high-volume inference. It's plausible that Nscale's architecture, heavily optimized for training, was less efficient for the inference workloads that were becoming the dominant cost center for services like Copilot. Microsoft's internal Maia chip, on the other hand, was reportedly being designed with a more balanced profile, better suited for this new reality. The initial synergy that made the partnership so compelling began to erode as the problem they were trying to solve changed beneath their feet.
The Official Announcement: Deconstructing Microsoft's Public Statement
Corporate announcements about terminated partnerships are masterpieces of carefully crafted language, designed to minimize market panic and legal exposure. Microsoft's statement on the Nscale agreement was no exception. It was a study in strategic ambiguity, and deconstructing its key phrases reveals much about the underlying reality. The press release led with the line that both companies had 'mutually agreed to wind down their co-development and supply agreement.' The phrase 'mutually agreed' is standard corporate-speak for a separation that is anything but. It implies a collaborative decision, but in a relationship with such a profound power imbalance, it almost certainly means Microsoft made the decision and Nscale was forced to accept it. The use of this phrase is intended to present a united front and avoid the appearance of a contentious breakup that could lead to protracted legal battles.
The statement cited 'evolving strategic priorities and a regular review of our technology roadmap' as the primary driver for the change. This is the core of the message, and it is deliberately vague. 'Evolving strategic priorities' points directly to Microsoft's increasing focus on its own custom silicon. It allows Microsoft to frame the decision not as a failure of Nscale's technology, but as a proactive choice to vertically integrate a critical component of its infrastructure. This narrative is crucial for maintaining investor confidence in Microsoft's AI strategy. It suggests foresight and control, rather than a reactionary move to ditch a failing supplier. The 'regular review of our technology roadmap' clause serves to normalize the decision, positioning it as a routine part of business rather than an emergency course correction.
Notably absent from the announcement was any specific criticism of Nscale's hardware or execution. Microsoft's statement included platitudes, thanking the Nscale team for their 'innovative contributions and collaboration.' This is a standard professional courtesy designed to soften the blow and avoid burning bridges entirely. Publicly disparaging a partner's technology could expose Microsoft to legal claims and damage its reputation as a reliable partner for other startups. By keeping the language positive, Microsoft leaves the door open for future, smaller-scale collaborations and signals to the rest of the industry that it can be a constructive, if demanding, partner. The message is clear: the partnership ended not because Nscale failed, but because Microsoft's own strategy evolved beyond it. This carefully constructed narrative places the onus of the change squarely on Microsoft's own internal calculus, shielding it from accusations of having picked the wrong horse in the AI hardware race.
Nscale's Response and the Immediate Fallout
For Nscale, Microsoft's announcement was a cataclysmic event. While Microsoft's statement was couched in neutral corporate language, Nscale's response was, by necessity, a desperate exercise in damage control. Their press release, issued hours after Microsoft's, attempted to put a brave face on the situation. It emphasized the 'maturity and performance' of their TFU technology and stated that the end of the Microsoft agreement would allow them to 'accelerate engagement with a broader set of customers across the cloud and enterprise markets.' They highlighted their intellectual property portfolio and the expertise of their engineering team as key assets that would ensure the company's long-term success. However, the market's reaction was swift and brutal.
If Nscale had been a publicly traded company, its stock would have plummeted by 80-90% within minutes. As a private, venture-backed entity, the fallout was less visible but no less severe. Its valuation, which had been in the tens of billions based on the Microsoft deal, evaporated overnight. Subsequent funding rounds became nearly impossible, as investors questioned the company's ability to survive without its anchor tenant. The most immediate and painful consequence was massive layoffs. With the primary source of its future revenue gone, Nscale was forced to slash its workforce dramatically, reportedly letting go of over 60% of its staff, including sales, marketing, and the large engineering teams dedicated specifically to the Microsoft integration. This not only crippled the company's operational capacity but also triggered a crisis of morale among the remaining employees, leading to a wave of voluntary departures of top talent who saw the writing on the wall.
Customers and potential customers who had been in talks with Nscale immediately paused all engagement. The implicit promise of Nscale's technology was that it was 'Microsoft-validated.' The cancellation acted as a vote of no confidence from the one customer that mattered most. Enterprises that had been considering Nscale for their own private AI clouds were now deeply skeptical. Why should they bet their infrastructure on a technology that the world's leading software company had just abandoned? The fallout demonstrated the immense risk of the 'single-customer' business model. Nscale had bet its entire existence on Microsoft, and when that bet failed, it was left with a world-class technology but no clear path to market. The company was thrown into a desperate fight for survival, forced to pivot its entire business strategy in a matter of weeks while dealing with the financial and human toll of the collapse.
The Ripple Effect: Impact on the AI and Data Center Industry
The termination of the Microsoft-Nscale agreement sent ripples far beyond the two companies involved, triggering a significant re-evaluation of strategy across the AI hardware and data center sectors. For other AI chip startups, the news was a chilling reminder of the dangers of relying on hyperscale customers. Venture capital firms, which had been pouring billions into companies hoping to become the 'next Nvidia,' suddenly became much more cautious. The Nscale story became a cautionary tale, and the bar for new investments in AI hardware startups was raised significantly. VCs began prioritizing companies with a more diversified customer strategy, targeting a mix of enterprise, research, and smaller cloud providers, rather than hunting for a single transformative deal with a tech giant.
For Microsoft's main competitors, AWS and Google Cloud, the announcement was a moment of strategic validation. Both companies had long since committed to developing their own custom silicon: AWS with its Trainium and Inferentia chips, and Google with its Tensor Processing Units (TPUs). Microsoft's retreat from a major third-party supplier and its renewed emphasis on its in-house 'Maia' project affirmed that vertical integration is the winning strategy for controlling performance, cost, and supply chain at hyperscale. It signaled that the era of cloud providers acting as mere integrators of third-party hardware was ending. To compete at the highest level of AI, owning the entire stack, from the silicon to the software, is becoming non-negotiable. This likely accelerated investments in their own chip design teams and hardened their resolve to rely on internal solutions for their most critical workloads.
Nvidia, the dominant player in the AI accelerator market, emerged as an indirect winner. The failure of a major potential competitor reinforced Nvidia's market position. For enterprise customers and smaller cloud providers, the Nscale collapse made Nvidia's CUDA platform seem like an even safer, more stable bet. The event highlighted the immense difficulty of building not just a competitive chip, but also the entire software ecosystem required to make it useful. Nvidia's decades of investment in CUDA created a deep moat that Nscale, even with Microsoft's backing, could not cross quickly enough. The data center industry also took note. Companies involved in building and equipping facilities, like colocation providers and equipment manufacturers, saw this as a sign of increasing concentration of power. The custom designs of hyperscalers mean less standardization, requiring more bespoke solutions for cooling, power delivery, and server rack design, which benefits some suppliers while marginalizing others who rely on high-volume, standardized components.
Technical Debt or Strategic Pivot? Analyzing the Root Causes
Dissecting the failure of the Nscale partnership requires looking beyond the official statements to the complex interplay of technology, strategy, and timing. The central question is whether the deal collapsed because Nscale's technology failed to deliver (a 'technical debt' problem) or because Microsoft's strategic needs changed (a 'strategic pivot'). The reality is likely a combination of both, with each factor reinforcing the other.
Evidence for the 'technical debt' argument centers on execution risk. Developing a new chip architecture from scratch is monumentally difficult. Reports of missed performance targets on key internal Microsoft benchmarks suggest that Nscale's real-world results may not have matched the promises made on paper. This is a classic challenge: a design that is theoretically superior can be hampered by practical issues in the compiler, firmware, or manufacturing process. For example, while Nscale's 'Synapse Link' interconnect was brilliant in theory, it may have suffered from software maturity issues, making it difficult for Azure's engineers to achieve the promised 'single logical accelerator' experience. If Microsoft's internal teams found themselves spending more time debugging Nscale's stack than developing new AI models, the perceived value of the partnership would rapidly diminish. Every quarter that Nscale's hardware was delayed or underperformed, the case for Microsoft's own 'Maia' chip grew stronger.
Conversely, the 'strategic pivot' argument is equally compelling. Microsoft's AI strategy is not static; it evolves with the market. The explosion of generative AI and the success of products like GitHub Copilot shifted the economic calculus from pure training performance to the total cost of ownership (TCO) across both training and inference. A chip that is the absolute fastest at training a model once might be economically unviable if it's inefficient at serving millions of inference requests per day. Microsoft's deep operational data from running Azure and its own AI services gave it a unique insight into these evolving workload patterns. This is a journey many data teams experience, which is why understanding modern platform shifts, like the move detailed in Power BI Premium retirement and Microsoft Fabric migration, is so crucial for data professionals. It's highly probable that Microsoft's internal analysis revealed that its own 'Maia' architecture, co-designed with its software and services teams, could offer a better TCO for the mixed workloads they anticipated in 2026 and beyond. In this view, Nscale's technology wasn't necessarily 'bad,' but it was optimized for a problem that was no longer Microsoft's most important one. The cancellation was a forward-looking decision to align its hardware investments with its future, inference-heavy service needs, rather than being locked into an architecture optimized for the past.
Lessons for Enterprise Partnerships in the AI Era
The dramatic collapse of the Microsoft-Nscale deal offers a powerful set of lessons for technology companies, from startups to established enterprises, about the nature of strategic partnerships in the fast-moving AI sector. These cautionary tales highlight the risks of dependency, the importance of aligned incentives, and the need for constant strategic re-evaluation.
The first lesson is for startups: the danger of the 'whale hunter' strategy. Landing a single, massive customer like Microsoft can seem like the ultimate victory, providing validation, revenue, and scale. However, it creates a fragile business model where the startup's fate is entirely tied to the strategic whims of its one client. Nscale's experience demonstrates that a startup in this position has very little leverage. When the larger partner's priorities shift, the smaller company is left with little recourse. Future AI hardware and software startups should learn from this and prioritize building a diversified customer base from the outset, even if it means slower initial growth. Having multiple, smaller customers creates a more resilient business that is not subject to the existential threat of a single contract cancellation.
For large enterprises like Microsoft, the lesson is about the true cost of 'buying' innovation. While partnering with a nimble startup can accelerate development in the short term, it also introduces significant integration risk and a loss of direct control over a critical part of the technology stack. The time and resources Microsoft spent trying to integrate Nscale's technology, only to abandon it, represent a massive opportunity cost. This experience reinforces the argument for vertical integration in core strategic areas. If a technology is truly fundamental to a company's future, the long-term benefits of developing it in-house, where it can be perfectly tailored to the company's needs and roadmap, often outweigh the short-term speed advantages of an external partnership. This is a recurring theme in Microsoft's product strategy, seen even in how it evolves its developer tools, where choices like those in a Botpress vs Rasa vs Microsoft Bot Framework comparison reflect a constant re-evaluation of internal versus external technology.
Finally, for any company entering into a deep technical partnership, the importance of continuous and transparent communication about roadmaps cannot be overstated. The Nscale deal likely soured in part because the two companies' long-term visions began to diverge. A successful partnership requires mechanisms for regular, high-level strategic reviews to ensure that both parties are still working towards the same goals. Contracts should also include clearer off-ramps and contingency plans. Relying on goodwill is not enough; partnerships need to be structured to withstand the inevitable shocks and changes of a dynamic market like AI. The Nscale story is a stark reminder that in technology, alignment is not a one-time event at the signing of a deal, but a continuous process of recalibration.
The Future of Nscale: Survival, Acquisition, or Obscurity?
With the Microsoft lifeline severed, Nscale faces a daunting future with three potential paths: a strategic pivot to survive as an independent company, an acquisition by another large technology player, or a slow slide into obscurity and an eventual asset sale. Each path presents its own monumental challenges and slim chances of success. The path of independent survival is the most difficult. To succeed, Nscale must rapidly execute a dramatic pivot. Its new strategy would need to focus on a market segment that the hyperscalers and Nvidia are not directly targeting. This could include specialized high-performance computing (HPC) for scientific research, sovereign AI clouds for national governments, or on-premises solutions for large enterprises in finance or pharmaceuticals with strict data privacy requirements. This requires a complete retooling of its sales and marketing efforts and a product that can be easily adopted without the massive engineering support of a partner like Microsoft. The company would need to raise a significant new round of funding in a skeptical market, a herculean task given its recent history.
The second, and perhaps more likely, positive outcome is acquisition. While Microsoft has walked away, Nscale's intellectual property and its core engineering team still hold immense value. A rival cloud provider, such as Oracle or a large Asian cloud company, might see an opportunity to acquire Nscale's technology at a steep discount and jumpstart its own custom AI hardware efforts. A large semiconductor company, like Intel or AMD, could also be a potential suitor, looking to integrate Nscale's architecture into its own portfolio to better compete with Nvidia. A major enterprise technology firm, like Dell or HPE, might see Nscale as a key component for building turnkey private AI cloud solutions. However, any acquisition would be complex. The acquirer would need to be confident that it could succeed where Microsoft failed, and the acquisition price would be a fraction of Nscale's former valuation, a bitter pill for its early investors and employees to swallow.
The third and most tragic path is a slow decline into irrelevance. If Nscale is unable to secure new funding or a buyer, it would be forced into a wind-down scenario. This would likely involve further layoffs, scaling back operations to a skeleton crew focused on servicing any remaining small contracts, and eventually selling off its patents and other assets to the highest bidder. This is the fate of many promising hardware startups that fail to achieve commercial escape velocity. The technology itself might be brilliant, but without a viable market and a sustainable business model, even the most innovative designs can end up as little more than a collection of patents in a larger company's portfolio. The ultimate fate of Nscale will serve as a key data point for the next generation of AI hardware entrepreneurs and investors.
Microsoft's Path Forward: Doubling Down on In-House AI Infrastructure
For Microsoft, the cancellation of the Nscale agreement is not a setback but a strategic clarification. It marks a definitive turn away from reliance on external partners for core AI hardware and a full-throated commitment to vertical integration. The future of Azure's AI infrastructure now rests squarely on the shoulders of its own internally designed silicon, primarily the Maia series of accelerators and the Athena networking processors. This strategy offers Microsoft several compelling long-term advantages: cost control, performance optimization, and supply chain security.
By designing its own chips, Microsoft can tailor the hardware precisely to the needs of its software stack and its most important workloads, such as running large language models for Copilot and Azure OpenAI services. This co-design process allows for optimizations that are impossible when using off-the-shelf components. Microsoft can build features directly into the silicon that accelerate its specific software frameworks, leading to better performance and lower operational costs. This control over TCO is paramount at the scale Microsoft operates; even a small percentage improvement in efficiency translates into billions of dollars in savings. This focus on building a cohesive internal platform is a consistent theme, impacting decisions around their entire infrastructure, including their extensive network of data centers, such as the major Microsoft data centre development in Essex, which forms part of this broader self-reliant strategy.
Furthermore, an in-house strategy insulates Microsoft from the notoriously volatile semiconductor supply chain. By managing its own design and working directly with foundries like TSMC, Microsoft reduces its dependency on a single supplier like Nvidia, whose GPUs have often been in short supply. This gives Microsoft more predictable control over its own destiny, allowing it to scale its AI infrastructure in line with its own ambitious growth plans, rather than being constrained by the production roadmap of a third party. The move also enhances its competitive posture. As AI becomes more commoditized, the efficiency of the underlying infrastructure will be a key differentiator. Owning the entire stack allows Microsoft to compete more effectively on price and performance against AWS and Google, who have long pursued similar strategies. The Nscale chapter will ultimately be remembered as a transitional period, a costly but necessary detour on Microsoft's journey to becoming a truly self-sufficient AI superpower.
Implications for AI Professionals and Skills Development in 2026
The strategic shifts in the hardware layer, exemplified by the Microsoft-Nscale split, have profound implications for the skills that will be most valuable for AI professionals in 2026. The era of simply writing Python code against a generic GPU in the cloud is evolving. As hyperscalers like Microsoft, Google, and AWS move toward highly specialized, vertically integrated stacks, engineers who can operate effectively at the intersection of hardware and software will be in extremely high demand. This is a key focus at organizations like Refonte Learning, which recognizes that future-proof skills require a deeper understanding of the full system.
First, expertise in performance engineering and optimization will become critical. When a company owns the entire stack from silicon to application, it creates enormous opportunities for cross-layer optimization. AI engineers and data scientists will need to move beyond just tweaking model architectures and hyperparameters. They will need to understand how their code interacts with the underlying custom hardware. This includes skills in using specialized compilers, profiling tools to identify hardware-specific bottlenecks, and techniques like quantization and model pruning that are tailored to the unique characteristics of a chip like Microsoft's Maia. The ability to squeeze maximum performance out of a given piece of silicon will directly impact the profitability of AI services, making these skills incredibly valuable.
Second, a deep understanding of distributed systems remains non-negotiable, but with a new twist. The challenge is no longer just about scaling out across a cluster of identical nodes. Instead, it involves orchestrating complex workflows across heterogeneous hardware environments that might include a mix of CPUs, GPUs, and multiple generations of custom AI accelerators. Professionals who have mastered frameworks for distributed computing like Ray, and who understand the networking principles behind technologies like Microsoft's Athena, will be essential for building and maintaining the next generation of AI platforms. This requires a systems-level thinking that transcends a single programming language or framework. Aspiring professionals looking to build these durable, in-demand capabilities should explore comprehensive training like the AI Engineering Program, which is designed to cultivate this deep, systems-oriented expertise.
Finally, adaptability and a commitment to continuous learning will be the most important soft skills. The hardware landscape is becoming more fragmented and is changing faster than ever. An engineer who is an expert on one proprietary stack today may find that stack obsolete in three years. The most successful professionals will be those who focus on a strong foundation in computer science fundamentals: algorithms, data structures, computer architecture, and operating systems. These first principles will allow them to adapt quickly as new hardware and software platforms emerge. The Microsoft-Nscale story is a clear signal that the AI industry is not settling down; it is entering a new phase of intense innovation and competition at the hardware level, creating immense opportunities for those with the right foundational skills.
Conclusion: A Strategic Uncoupling with Long-Term Consequences
The cancellation of the Microsoft-Nscale agreement is more than the failure of a single business deal; it is a defining moment in the evolution of the AI industry. It represents the logical conclusion of the trend towards vertical integration, where the world's largest technology companies are choosing to build, not buy, the foundational components of their AI future. This strategic uncoupling marks the end of an era of symbiotic partnership between hyperscalers and specialized hardware startups and ushers in an age of intense, direct competition at the silicon level.
For the broader ecosystem, the consequences will be felt for years. The venture capital landscape for AI hardware has been reshaped, with a new premium placed on diversified go-to-market strategies and a healthy skepticism towards any startup reliant on a single, massive customer. For Microsoft's competitors, it is a powerful affirmation of their own in-house silicon strategies. For Nvidia, it solidifies its market dominance by demonstrating the immense difficulty of building a viable alternative ecosystem from scratch. And for Nscale itself, it is a harsh lesson in the perils of flying too close to the sun, a cautionary tale that will be studied in business schools and boardrooms for the foreseeable future.
Looking ahead to 2026, the battle for AI supremacy will be fought not just with algorithms and data, but with custom-designed chips and tightly integrated infrastructure. Professionals who wish to thrive in this new environment must cultivate a deeper, more holistic skill set that bridges the gap between software and hardware. As platforms become more proprietary and complex, the need for robust, foundational training has never been greater. For those committed to leading in this new era, developing expertise in systems-level performance engineering and distributed computing is no longer optional; it is the critical path to success. Programs like the AI Engineering Program offered by Refonte Learning are designed specifically to equip the next generation of engineers with the durable, fundamental skills required to build and scale the future of artificial intelligence on this rapidly evolving landscape.