By mid-2019, the industry had industrialized, and with it came a critical realization: the bottleneck was no longer just access to electricity, but the reliability and efficiency of the infrastructure that consumed it. This was the backdrop for a presentation that sought to reframe the role of the data center not merely as a landlord of sockets, but as an active participant in the economic engine of the blockchain.
On July 23–25, 2019, at Mining Disrupt in Miami, Kristy-Leigh Minehan, then Core Scientific's chief technology officer, presented the company's strategy for bridging the gap between raw power and profitable mining. The company said it aimed to become the premier infrastructure and software provider for blockchain and artificial intelligence. Their mission was clear: to build a diversified, sustainable business that focused on infrastructure applications and digital asset mining. For a company that had been operating in stealth mode for roughly a year, this was the moment they lifted the veil, revealing not just their power capacity, but the software architecture that would define their competitive advantage.
The Infrastructure Gap: Why Miners Need More Than Sockets
Core Scientific argued that the traditional model of colocation was fundamentally misaligned with the needs of the blockchain ecosystem. In a symbiotic relationship between hardware vendors, infrastructure providers, and miners, every participant enters a social contract with the network. The infrastructure provider’s job is not simply to deliver power, but to act as the reliable backbone that ensures miners can continue to deploy and reinvest. The company noted that Satoshi Nakamoto himself recognized the driving force behind professional server farms back in 2010, predicting that many nodes would end up being large consumer server farms. However, the modern data center had to evolve to meet the demands of a market where downtime is not just an inconvenience, but a direct loss of revenue.
To address this, Core Scientific identified three key areas necessary for infrastructure to succeed in the blockchain space. The first was a fundamental shift in how hosting rates were structured. By moving away from flat-fee models, the company sought to align its financial incentives with the actual performance of the customer’s hardware. This approach was designed to ensure that the data center’s success was inextricably linked to the profitability of the miners it served, creating a flywheel of success where network participation drives value, and value drives further participation.
The Social Contract: Aligning Incentives Through Kilowatt-Hours
The core of this alignment was a simple but powerful economic mechanism. Core Scientific stated that they do not recognize revenue unless the customer's machine is up and mining. This contractual arrangement meant that the infrastructure provider was financially motivated to keep the hardware running at peak efficiency, transforming the relationship from a passive lease to an active partnership. It was a recognition that in a competitive mining environment, the cost of power is only one component of the equation; the cost of inefficiency, downtime, and suboptimal performance is often just as significant.
an infrastructure provider is a service provider by tying our hosting rate to kilowatt an hour
Core Scientific · 6:17
Building Minder: The Software Layer for Operational Excellence
The second pillar of Core Scientific’s strategy was the recognition that infrastructure is nothing without optimization. Colocation is only valuable if it allows miners to make more money than they would by hosting at home. To achieve this, the company argued that optimization must take place at every point in the stack, from the mining pools to the physical power supply units. However, a significant gap existed in the market: most monitoring and management tools were built by miners for miners, lacking the enterprise-grade robustness and compliance features required by large-scale data centers.
Core Scientific realized that to be a world-class infrastructure provider, they needed to build the tools themselves. This led to the development of Minder, a software platform designed to automate tasks, reduce operational expenditure, and ensure a low cost footprint. Minder was built to track devices throughout their entire lifecycle, from unpacking and onboarding to warranty repairs and returns. It allowed the company to handle revenue recognition in line with standard accounting practices and remain compliant with auditing frameworks recognized by major firms like Deloitte, EY, PwC, and KPMG. For data center technicians, Minder provided a single pair of eyes and ears within the facility, managing both sides of the business from the moment a device’s shipping label was scanned to its final deployment.
we realized if we wanted to be a world-class infrastructure provider we needed to build the tools
Core Scientific · 8:06
From Detection to Action: AI in the Data Center Loop
A critical differentiator for Minder was its ability to turn data into actionable events. It was not enough to simply have insight into hashrate; the system needed to detect events before they happened. Minder’s environmental health reports helped gain insights into leading indicators for machine failure or degradation by correlating shares, hardware errors, and temperature. This allowed the company to identify specific problem areas and ensure accountability across every layer of the mining operation. The platform’s ticketing system was at the heart of this process, allowing facility managers to track the root cause for failure and monitor the KPIs of their operations team.
The integration of artificial intelligence took this a step further with Minder’s auto-healing technology. The system would begin to troubleshoot a machine and attempt to self-heal if it ran into a problem. If it could not solve the issue, it would reach for the nearest human on call via text or phone. If the ticket was not answered within a reasonable timeframe, it would escalate through the chain of command. The system also learned from previous behaviors and patterns, keeping a track record of how humans had interacted with similar machines, thereby improving its diagnostic capabilities over time.
The Chip-Level Revolution: Tuning Hardware in Real-Time
Beyond operational stability, Core Scientific highlighted that AI also drives profit optimization. The company argued that driving profit is not about figuring out what to mine, but rather about figuring out when to mine. Minder combined statistical analysis with sentiment analysis to deliver uplift to mining operations. By understanding where market makers sit and the signals that drive profitability across multiple coins, the system could correlate Bitcoin price dips with monthly electricity bills to optimize switching strategies. This was managed on a machine level, allowing for granular control over when to switch coins based on real-time profitability signals.
The most sophisticated aspect of Minder was its machine tuning platform, which functioned in both automated and manual modes. In automated mode, the system sought to tune individual devices based on silicon leakage qualities, age, environmental factors, and the specific coin being mined. In manual mode, it enabled operators to have full control over their devices, eliminating the need for third-party tools or custom firmware to overclock machines. The platform broke down machines by part, allowing technicians to isolate hash boards and control boards in a context-friendly interface. This allowed for per-chip level optimization and provided embedded engineers with data sheets in-line, enabling them to view specifications without leaving the application.
driving profit is actually about isn't about figuring out what to mine it's figuring about when to mine
Core Scientific · 11:42
The Q4 2019 Promise: Democratizing Enterprise Tools
The backend of this ecosystem was Minder OS, an optimization stack featuring custom drivers, firmware, and tooling. Core Scientific claimed that Minder OS could drive profit uplift with up to a 20% hashrate increase and up to a 30% power decrease. The software was built to an enterprise standard for security and was available to all Core Scientific hosted customers as part of their value-add service. For GPU miners, Minder OS enabled the detection of mining algorithms and the modification of memory timings on the fly, with APIs that fed directly into the kernel. This allowed for the ability to modify timings according to each roll of the compute unit, a level of precision previously reserved for high-end engineering teams.
custom drivers custom firmware custom tooling to drive profit uplift with up to 20% hashrate increase
Core Scientific · 14:08
Seven Years Later: From Bitcoin Miner to AI Landlord
The company that took the Miami stage in 2019 did not stay a Bitcoin hosting business. Core Scientific filed for Chapter 11 in December 2022, after Bitcoin's slide and heavy debt, and emerged in January 2024. What carried it out was the idea Minehan sketched on stage: a data center sells performance and uptime, not just sockets. The customer that bought into it was CoreWeave, the AI cloud provider, which leased Core Scientific's power for GPUs.
On July 7, 2025, CoreWeave agreed to buy Core Scientific outright in an all-stock deal worth about $9.0 billion, taking control of about 1.3 GW of gross power and ending more than $10 billion of future lease payments. The deal never closed. On October 30, 2025, Core Scientific's stockholders voted it down; its largest shareholder, Two Seas Capital, argued the price undervalued the company, and the proxy adviser ISS backed the opposition. Core Scientific stayed independent and said it would convert most of its facilities to AI workloads.
It has done exactly that. In the first quarter of 2026, colocation became its largest business, at $77.5 million in revenue, while it sold 2,385 Bitcoin for $208.3 million to fund construction. CoreWeave's contracts cover about 590 MW and $10.2 billion over 12-year terms. In July 2026 the company paid $41.9 million to cancel a Bitcoin miner order and said it is winding down mining; in the second quarter, colocation brought in $136.7 million, 83% of revenue, all of it from CoreWeave. A new set of agreements with AMD covers about 530 MW and more than $14 billion over 15 years. The miner that once promised to tune every hash board now rents power to AI.
Looking back from 2026, the argument made by Core Scientific has aged with the industry. When the session was recorded, the price of Bitcoin was $9,770. Today, the price stands at $82,880, reflecting a massive expansion of the network’s value and the corresponding industrialization of mining. The focus on reliability and software-driven optimization, once a novel proposition, has become the baseline expectation for any serious infrastructure provider. The conversation continues at Intelligence Disrupt 2027, March 22 to 24 in Irving, Texas, where the next generation of infrastructure challenges will be debated. The shift from selling power to selling performance has defined the years since.