Microsoft Discovery: Advancing agentic R&D at scale

Expanded preview entree for Microsoft Discovery brings caller enterprise-grade, agentic AI capabilities for probe and improvement teams.

Over the past year, we’ve made important advancement with Microsoft Discovery by moving intimately with probe and improvement (R&D) organizations. Today, we’re sharing however those efforts are translating into existent momentum for customers and partners, portion besides expanding preview entree to Microsoft Discovery. This adjacent signifier reflects what we’ve learned arsenic we proceed to broaden entree to enterprise-grade, agentic AI capabilities for R&D. The Microsoft Discovery level continues to germinate with caller capabilities, expanded spouse interoperability, and a increasing acceptable of results with real-world technological outcomes and engineering transformation. We judge what comes adjacent tin meaningfully alteration however R&D teams run and empower them to execute more.

The epoch of agentic AI for probe and development 

Agentic AI opens a caller section for R&D wherever autonomous cause teams, guided by quality expertise, execute the halfway probe and engineering tasks successful a redefined agentic loop. Specialized agents tin crushed connected apical of immense amounts of organizational and public-domain knowledge, make hypotheses connected an expanded hunt space, trial and validate those hypotheses astatine scale, analyse the results, and provender conclusions into iterative loops. Empowering subject and engineering experts with agentic AI has the imaginable to reshape the aboriginal of subject and engineering, enabling organizations to pb boldly successful the caller Frontier R&D era.

This cardinal displacement requires a heavy translation that encompasses some technological and organizational challenges. Scientific find has ever been defined by ambition and the relentless pursuit of what comes next—a much sustainable material, a cleaner root of energy, a much effectual treatment. But for galore R&D teams the hardest enactment tin statesman aft an thought shows promise. Turning concepts into outcomes requires repeated improvement cycles that impact reformulating candidates arsenic caller datasets emerge, re-engineering existing materials to conscionable evolving regulatory and show requirements, oregon adjusting designs erstwhile performance, yield, oregon manufacturability autumn short. As R&D grows much complex, tooling indispensable germinate to assistance adjacent the region betwixt what researchers and engineers privation to prosecute and what they tin practically deliver.

Earlier generations of AI offered incremental alleviation done faster hunt and amended retrieval, but lacked the deeper reasoning that genuinely complex, multi-disciplinary subject demands. Tradeoffs crossed cost, performance, yield, compliance, and timelines indispensable beryllium revisited repeatedly arsenic improvement progresses. But the convergence of large-scale reasoning models, agentic AI architectures, and high-performance unreality infrastructure has created a genuine accidental to rethink however R&D enactment gets done—not lone to amended existing processes astatine the margins, but to assistance teams iterate faster and determination from proposal to campaigner improvement to result with greater confidence.

Figure 1

When Microsoft Discovery was introduced successful backstage preview past year, it was an aboriginal look of that possibility: an agentic AI level purpose-built for R&D, bringing unneurotic the reasoning extent and collaborative quality that complex, real-world R&D requires. The effect from engineers and researchers crossed beingness sciences, chemistry and materials science, physics, semiconductors, and different fields made wide that the request was existent and the attack was right.

The Microsoft Discovery platform 

Microsoft Discovery is an extensible level that brings unneurotic agentic orchestration, precocious reasoning, a graph-based cognition foundation, and high-performance computing. It helps thrust the 3 principles outlined successful Figure 1 for effectual agentic discovery—enabling cause empowerment, find loop automation, and prime astatine scale. Because it is built connected Microsoft Azure’s endeavor unreality infrastructure, Microsoft Discovery is designed to run wrong the security, compliance, transparency, and governance frameworks utilized to negociate delicate real-world R&D environments.

Figure 2

Agents are equipped with a wide scope of digital, physical, and analytical tools utilized crossed R&D. This includes successful silico experimentation environments specified arsenic high-performance compute (HPC) clusters, specialized ample quantitative models (LQMs) and agents, and imaginable aboriginal integration with quantum capabilities arsenic they go applicable to commercialized R&D. It besides allows interoperability with carnal labs, facilitating the laboratory process procreation and adjacent nonstop cognition with robotics, laboratory instrumentation, and Internet of Things (IoT)-enabled devices that agents tin run nether quality oversight.

At the bosom of Microsoft Discovery is the Discovery Engine that mimics the technological method wherever specialized agents crushed over large amounts of knowledge, make hypotheses, and validate them in a analyzable histrion crossed a immense hunt space. The Discovery Engine connects proprietary probe information with outer technological literature—not solely to retrieve isolated facts but to crushed crossed conflicting theories, experimental results, and domain-specific assumptions successful a mode that reflects however science actually works. This contextual extent is what separates Microsoft Discovery from general-purpose AI tools and enables the level to relation arsenic a genuine reasoning spouse crossed the afloat arc of a probe program.

Built-in governance controls assistance guarantee that cause driven probe remains aligned with strategical priorities, information and compliance standards, and information requirements. These systems supply centralized management, audit trails, and checkpoints that assistance support reliability arsenic agentic throughput grows. The level is extensible by plan which enables integration with existing concern tools and assets, spouse solutions, and open-source models. Integration with Microsoft 365, Microsoft Foundry, and Microsoft Fabric enables organizations to interoperate crossed concern agents, endeavor data, and organization knowledge.

Real-world interaction of Microsoft Discovery 

Previously we shared how a squad of Microsoft researchers leveraged precocious AI models and HPC tools from Microsoft Discovery to place a novel, non-PFAS, immersion datacenter coolant prototype successful astir 200 hours. We’re excited to stock a fewer examples of however customers person been utilizing the level during preview.

Syensqo

A planetary person successful precocious materials and specialty chemicals, Syensqo is advancing a bold, multi-year translation of its exertion scenery to accelerate data-driven science, precocious simulation, and AI-enabled discovery. Building connected aboriginal occurrence with Microsoft Discovery, Syensqo is present scaling these capabilities enterprise-wide to unlock greater technological and concern impact. This adjacent signifier focuses connected modernizing R&D cognition foundations, expanding entree to scalable, cost-efficient, cloud-based compute, and establishing a unified operating exemplary that brings unneurotic data, high-performance computing, and emerging agentic AI to powerfulness the aboriginal of innovation.

As Microsoft Discovery workflows gained momentum, Syensqo expanded its ambition to standard these capabilities crossed some R&D and commercialized organizations, unlocking caller opportunities for end-to-end innovation. This improvement is enabling teams to unify technological and concern datasets, standard simulation environments successful enactment with progressively analyzable development needs, and integrate engineering workflows wrong a connected integer ecosystem. Together, these advancements are establishing a strong, future-ready instauration to accelerate innovation-led growth—from early-stage find done engineering and large-scale formulation. 

To recognize this vision, Syensqo is advancing its subject and commercialized information and simulation platforms connected Azure. By centralizing captious datasets wrong a governed, enterprise-grade information backbone and extending Microsoft Discovery workflows onto highly scalable unreality compute, the institution is establishing a modern, standardized operating exemplary for innovation. This displacement enables much seamless collaboration, supports precocious analytics and simulation astatine scale, and lays the groundwork for next-generation, AI-powered workflows crossed precedence probe and innovation (R&I) domains.

We are entering a caller signifier of our concern with Microsoft, focused connected scaling AI agents crossed research, income and selling to thrust near-term growth. By connecting lawsuit request to technological improvement and backmost to marketplace execution, agentic AI is enabling faster cycles, sharper prioritization, and tangible interaction connected gross maturation and concern performance.” 

—Mike Radossich, Chief Executive Officer (CEO), Syensqo

GigaTIME  

Modern oncology progressively depends connected knowing tumors not lone by appearance, but by the biologic signals that signifier compartment behavior, immune response, and attraction outcomes. GigaTIME addresses this request by utilizing AI to infer spatially resolved tumor microenvironment signals from regular hematoxylin and eosin (H&E) pathology slides. This attack makes insights specified arsenic immune infiltration, checkpoint context, and tumor proliferation much accessible astatine scale without the outgo and throughput constraints of experimental assays. GigaTIME and its outputs wrong Microsoft Discovery are intended for probe usage only. They are not a aesculapian instrumentality and are not intended for objective diagnosis, treatment, prevention, oregon patient-management decisions. 

The interaction of GigaTIME increases erstwhile its outputs are embedded into existent probe workflows. Within Microsoft Discovery, virtual multiplex immunofluorescence (mIF) predictions determination beyond standalone visualizations and go inputs to ongoing technological reasoning. Spatial phenotypes tin beryllium generated consistently crossed cohorts, localized to azygous compartment context, and connected to supporting grounds specified arsenic literature, biomarkers, and downstream endpoints. This allows researchers to construe results systematically, question assumptions, and refine biologic hypotheses implicit time.

Microsoft Discovery supports this enactment successful a mode that is reproducible, scalable, and governed extremity to end. GigaTIME tin beryllium utilized alongside further models, information sources, and tools wrong a shared situation that supports iteration, comparison, and validation. Rather than accelerating a azygous analytical step, Discovery supports a afloat find loop—where spatial biology informs hypotheses, hypotheses usher validation, and results provender the adjacent rhythm of learning with clarity and confidence.

Learn much astir the GigaTIME and Microsoft Discovery integration to spot however virtual mIF outputs are applied wrong Microsoft Discovery for oncology R&D.

PhysicsX

PhysicsX, a person successful physics AI for concern engineering and manufacturing, is partnering with Microsoft to bring agentic engineering into accumulation done Microsoft Discovery. At the halfway of this collaboration is the PhysicsX platform—combining Large Physics Models and AI-native workflows to present near-real-time simulation by inference crossed the afloat engineering lifecycle.

Integrated into Discovery’s agentic environment, the PhysicsX level enables engineers to determination beyond sequential, solver-driven workflows and research importantly larger plan spaces, evaluating thousands of manufacturable candidates successful days, without compromising carnal fidelity.

The collaboration is already delivering interaction astatine Microsoft Surface. Faced with tightly coupled constraints crossed thermal performance, acoustics, and signifier factor, the Surface engineering squad utilized the PhysicsX level done Discovery to reimagine their cooling instrumentality plan process. What antecedently required weeks of simulation and manual setup is present compressed into days. Discovery agents orchestrate the generation, evaluation, and optimization of thousands of geometries, surfacing high-performing, production-ready designs for validation.

The effect is simply a measurement alteration successful engineering productivity: faster iteration, broader design-space coverage, and much assured decision-making. The attack is present being extended crossed further components successful the Surface portfolio.

Engineering is inactive constrained by workflows built for the pre-AI era. This concern changes that. PhysicsX’s frontier physics AI models, combined with Microsoft Discovery’s agentic orchestration and Azure infrastructure, springiness engineers the quality to research plan spaces that were antecedently retired of reach—at the velocity and standard that modern concern improvement demands.

—Jacomo Corbo, CEO, PhysicsX

Synopsys

Synopsys is simply a person successful physics plan automation (EDA), machine aided engineering (CAE) tools, and intelligence spot (IP), and plays a cardinal role in the plan and improvement of the astir analyzable chips and systems for the starring semiconductor and systems companies of the world.  

Synopsys and Microsoft person been partnering since 2019, helping pioneer software-as-a-service (SaaS) models connected Microsoft Azure. Synopsys besides launched the archetypal Silicon Copilot successful collaboration with Microsoft and is continuing that travel by leveraging Microsoft Discovery to rotation retired solutions for spot design.

The semiconductor manufacture is facing an unprecedented acceptable of challenges—demand for precocious show chips is increasing exponentially, complexity of sustainable, power-efficient spot design, and a captious shortage of skilled engineering. Agentic systems tin assistance mitigate these challenges portion accelerating plan cycles.

Synopsys agentic AI stack with multi-agent workflows built connected AgentEngineer™ technology, supported by Microsoft Discovery, person defined a caller paradigm for the industry.

Chip plan sits astatine the intersection of utmost complexity and outsized impact—exactly wherever AI tin marque the biggest difference. By bringing unneurotic Synopsys’ AI‑driven plan enactment with Microsoft Discovery, we are enabling agentic AI to redefine semiconductor engineering workflows, unlock step‑function productivity gains, and accelerate the adjacent epoch of exertion innovation.

—Ravi Subramanian, Chief Product Management Officer, Product Management & Markets Group, Synopsys

A increasing ecosystem

Microsoft Discovery works with an expanding ecosystem of partners offering integrated tools and specialized expertise.

Expanding what is possible for R&D 

Expanding the preview marks an important step in making agentic AI disposable to a broader acceptable of R&D organizations. Microsoft Discovery reflects our content that the adjacent procreation of technological progress can come from systems that harvester human expertise with AI that tin reason, plan, and enactment astatine scale. 

We look guardant to partnering with organizations that privation to rethink however find happens and to assistance signifier the aboriginal of endeavor R&D. 

For organizations looking to get started with Microsoft Discovery be definite to review the technical documentation to recognize requirements, onboarding prerequisites, and infrastructure considerations.

Microsoft Discovery is offered in preview. Features, availability, integrations, and show characteristics described successful this station whitethorn alteration anterior to, oregon without, wide availability and are not commitments. Statements astir aboriginal capabilities (including immoderate imaginable quantum integration) are forward-looking and taxable to change. Customer and interior outcomes described bespeak circumstantial workflows and data; individual results volition vary.