Public cloud vs. on-prem: Summit on where each workload belongs

Key Concepts

• Public Cloud: Hyperscale infrastructure (AWS, Azure, etc.) offering rapid scalability and "Lego-like" building blocks.
• Private Cloud: Dedicated compute and storage resources, often collocated in a data center, providing more control and cost predictability.
• Hybrid Cloud: A strategic mix of public and private infrastructure tailored to specific workload maturity.
• Workload Maturity: The lifecycle stage of an application; early-stage/prototyping benefits from public cloud, while stable/mature workloads often benefit from private cloud.
• Data Sovereignty: The requirement to keep data within specific geographic or regulatory boundaries.
• Repatriation: The process of moving workloads from public cloud environments back to private or on-premises infrastructure.
• Compute Crunch: The global shortage of high-performance hardware (GPUs, high-bandwidth memory) driven by AI demand.

---

1. The Evolution of Cloud Strategy

The conversation highlights a shift in enterprise IT. While startups remain "default cloud" to focus on product development, larger enterprises are facing "cloud fatigue." This is characterized by rising costs, increased complexity, and the realization that the "everything-to-the-cloud" mandate of the last decade was not a one-size-fits-all solution. Byron Dill notes that many companies are now experiencing a "slow burn" of rising bills, often triggered by leadership changes or audits, leading them to re-evaluate their IT footprint.

2. The "Lego" Problem vs. Bespoke Infrastructure

• Public Cloud: Offers a vast array of "nerd knobs" and services. While flexible, it forces companies to adapt their workloads to the provider's feature updates and rigid service models.
• Private Cloud (Summit’s Approach): Focuses on the specific needs of the workload. Instead of forcing a company to fit into a pre-defined cloud service, Summit designs infrastructure around the application’s requirements, supporting platforms like VMware, Hyper-V, Proxmox, and Kubernetes.

3. Security and Risk Management

Security is no longer just about firewalls; it is about "knowing your environment."

• The "Server Hugger" Mentality: Some organizations prefer on-premises control for compliance and business continuity.
• Shared Responsibility: While public clouds offer massive security research teams, clients are increasingly demanding granular control over their own policies, procedures, and controls to meet specific regulatory or investor requirements.
• De-risking: The primary driver for moving to private cloud is often risk mitigation—ensuring that critical assets are running in a predictable, stable, and sovereign environment.

4. The AI Impact and Hardware Scarcity

The current global compute crunch, particularly for GPUs and high-bandwidth memory, has made hardware acquisition difficult.

• Strategic Bifurcation: Companies are increasingly using a hybrid model: keeping rapid prototyping and new AI models in the public cloud for flexibility, while moving stable, high-compute AI workloads to dedicated private GPU clusters to optimize ROI.
• Supply Chain: Suppliers are reporting sold-out hardware through 2027. Summit manages this by leveraging long-standing partnerships and a global IP network to secure capacity for clients.

5. Methodology: The "Digital Therapist" Approach

Summit employs a consultative framework for clients:

1. Workload Readiness Assessment: A deep dive into existing applications to determine if they are "right" for the cloud or better suited for private infrastructure.
2. Discovery: Asking the "why" behind the infrastructure. Byron Dill emphasizes that the goal is not to blindly move off the cloud, but to ensure the infrastructure matches the business value.
3. Partnership vs. Vendor: Unlike the anonymous support of hyperscalers, Summit emphasizes a "first-name basis" relationship where engineers understand the client's specific environment and tribal knowledge.

6. Notable Quotes

• "Security isn't just a... firewall... it's very much security is knowing your environment and what's going on inside of it." — Byron Dill
• "Don't just follow the herd. Don't go wildly to one extreme then the other. Think about the actual workloads in question." — Alex Wilhelm
• "Our best clients where we have the most success are where we are partners with the client, not just providing services or a vendor." — Paul Armanacus (quoted by Byron Dill)

7. Synthesis and Conclusion

The main takeaway is that diversification is the new standard for enterprise IT. Just as one would diversify an investment portfolio, companies should not put all their "eggs" in one basket (either all-public or all-private).

For companies spending over $1 million annually on cloud services, a "workload readiness assessment" is a valuable exercise. The goal is to move mature, stable workloads to private environments to gain cost predictability (often targeting 40% savings) and control, while maintaining a public cloud presence for agility and innovation. Ultimately, IT infrastructure should be a value driver, not a "boat anchor" that consumes all human resources just to keep servers running.