Tech Talk: Switchboard

For the first time in blockchain history, price discovery is moving on-chain—and Solana is leading the charge. At Breakpoint 2025, Switchboard CTO Mitchell Gildenberg (known as Dr. Blox) revealed how a convergence of oracle innovations and new block ordering mechanisms is enabling trading infrastructure that could rival centralized exchanges at the microsecond level.

Summary

After five years of building oracle infrastructure for Solana, Switchboard is pioneering a new era of "just in time" oracles that fundamentally change how blockchain applications access real-world data. The presentation traced the evolution from expensive push-based oracles—which once consumed 3-7% of Solana's total traffic—to highly efficient systems that can deliver multiple price updates within a single 200-millisecond slot.

The breakthrough comes through integration with Jito BAM (Block Auction Mechanism), which creates a verifiable marketplace for transaction ordering. This eliminates the long-standing "one slot problem" where validators could arbitrarily drop oracle updates that didn't benefit them. With verifiable block ordering, market makers running prop AMMs (proprietary automated market makers) can now push price updates at any point within a block, enabling microsecond-level price accuracy.

This shift has profound implications for DeFi's competitiveness. Traditional exchanges like the New York Stock Exchange operate at 50-microsecond latency. Until now, decentralized applications couldn't come close to this standard. Dr. Blox explained that with the new architecture, Solana applications can now target microsecond-level pricing—putting them in direct competition with centralized venues for the first time.

The evolution represents a complete paradigm shift: where oracle providers once spent $500,000 monthly just to keep prices updated, the new model only triggers oracle transactions immediately before trades execute, dramatically reducing costs and network spam while improving accuracy.

Key Points:

The Evolution from Push to Pull to Just-In-Time Oracles

Oracle technology has undergone three distinct generations since 2017. The push model era saw oracle networks constantly broadcasting price updates regardless of whether anyone needed them. On networks like Ethereum, this meant heartbeat intervals of 30 minutes to manage costs that could reach thousands of dollars daily in gas fees. This approach worked for overcollateralized lending protocols but severely limited what DeFi applications were possible.

The pull model emerged around 2022, shifting the responsibility of requesting price updates to users. This dramatically reduced waste but still required users to initiate the price fetch before transactions could execute. Switchboard's "on-demand" variation took this further by having oracles fetch fresh data at the moment of request rather than pulling from a cached layer. This enabled sub-five-second flows for perpetual exchanges—an improvement, but still not competitive with centralized alternatives.

The Rise of Prop AMMs and Their Infrastructure Demands

Prop AMMs (proprietary automated market makers) are transforming spot trading on Solana by having market makers push their own price quotes every single slot. Unlike traditional AMMs that derive prices algorithmically from pool ratios, prop AMMs reflect real-time market maker quotes, dramatically improving price discovery and reducing spreads.

However, this model is extremely resource-intensive. Each oracle update consumes 70,000 to 120,000 compute units per asset, and if every emerging prop AMM maintains this cadence, the network faces millions of updates per epoch just to support a handful of venues. Dr. Blox questioned whether this approach is sustainable as the ecosystem scales, highlighting the need for more efficient solutions.

Solving the One Slot Problem with Jito BAM

The "one slot problem" has been a fundamental barrier to oracle improvement. With a single leader processing transactions in each slot, validators have no incentive to include multiple oracle updates—they can simply cherry-pick the most advantageous price for their own trades and drop the rest. This caps meaningful price updates at one per slot regardless of how fast oracles can actually deliver data.

Jito's Block Auction Mechanism creates a marketplace for transaction ordering with verifiable guarantees about sequence. This prevents sandwiching attacks and ensures that multiple oracle updates within a single slot are processed fairly. For the first time, intra-slot price updates become meaningful—even though Solana slots are 200 milliseconds, prices can now be updated at any point within that window.

Switchboard's BAM Plugin for Efficient Oracle Delivery

Switchboard is releasing its own BAM plugin that allows oracle updates to be inserted directly into the block auction layer rather than competing in the general transaction mempool. This means price updates can be placed immediately before trades execute, verifiably and efficiently.

The approach eliminates wasteful spam while ensuring maximum timeliness. Market makers no longer need to broadcast continuous updates hoping they land in the right position—they can precisely coordinate oracle updates with the trades that need them. This creates a fundamentally new model for pricing oracle services based on the actual value they provide to specific transactions.

Real-World Applications Already in Production

Switchboard's permissionless platform is already powering significant real-world use cases. Kiel used Switchboard to establish the NAV (net asset value) pricing for their $2.5 billion fund for MakerDAO on Solana—demonstrating that the oracle infrastructure is trusted for institutional-scale asset management. Jupiter uses Switchboard for prediction market data via Cauchy data integration, and the platform now supports the emerging prop AMM ecosystem.

The ability to build custom oracle workflows means applications aren't limited to standard price feeds. Teams can create oracles for arbitrary data sources, transform information on-chain, and establish consensus mechanisms for virtually any type of off-chain data.

Facts + Figures

• Switchboard has been providing oracle services on Solana for five years, since 2020
• Push-based oracles once consumed 3-7% of Solana's total network traffic
• Some oracle platforms were spending $500,000 per month just to maintain price updates
• The New York Stock Exchange operates at approximately 50-microsecond latency for price updates
• Solana slots are 200 milliseconds, but with Jito BAM, oracles can update prices at any point within that window
• Switchboard has optimized oracle updates to approximately 20 compute units on-chain
• Per-slot oracle updates consume 70,000 to 120,000 compute units per asset
• Push-based oracles in 2017-2022 often had heartbeat intervals of 30 minutes
• Kiel used Switchboard for NAV pricing of their $2.5 billion MakerDAO fund on Solana
• Jupiter uses Switchboard for prediction market data via Cauchy data integration
• The pull oracle model enables sub-five-second complete flows for perpetual exchanges
• This marks Dr. Blox's fifth presentation at Breakpoint

Top Quotes

"When switchboard is created, we're enabling people to make all these new mechanisms for how to push these prices."

"Really, when we started Switchboard, there was only a single method that people really used to push information from off-chain to on-chain."

"Unsustainable for any platform that's pushing information, and also if nobody's using the information between requests, it's wasted money, and wasted lock space."

"If Solana or any decentralized system is going to compete with that, we need to be coming to that as our Golden Bar or Golden Standard."

"Now with Jito BAM, we can make a platform for people actually executing oracle updates just before trades, to be as efficient as possible, as timely as possible."

"Now it's very interesting because now people that are quoting prices might look on-chain or on Solana first before even a CEX."

"With the advent of prop AMMs, price discovery is for the first time ever moving on-chain rather than off-chain."

Questions Answered

What is an oracle and why do blockchains need them?

An oracle is essentially a consensus mechanism for information. Blockchains operate in isolated environments without native access to external data like prices, weather, or website content. When you need to verify that a price from the internet is honest and legitimate, there's no built-in way for blockchain nodes to agree on that information. Oracles solve this by establishing trusted methods to fetch, verify, and deliver off-chain data to smart contracts. Switchboard built a platform where developers can create custom oracle workflows for any data source, rather than being limited to pre-defined feeds.

How did oracle technology evolve from push to pull models?

The push model dominated from 2017 to 2022, where oracle networks constantly broadcast price updates to the blockchain regardless of demand. This was extremely expensive—especially on Ethereum where gas costs could reach thousands of dollars daily. To manage costs, updates only occurred every 30 minutes, which limited DeFi to applications like overcollateralized lending. The pull model shifted responsibility to users, who request price updates only when needed. This eliminated wasted updates but added latency. Switchboard's on-demand model further improved this by having oracles fetch fresh data at the moment of request rather than from a cache.

What is the "one slot problem" and how does Jito BAM solve it?

The one slot problem refers to the fundamental conflict of interest when a single validator controls transaction ordering in a slot. If market makers are sending multiple price updates, validators have every incentive to cherry-pick the most advantageous one for their own trades and drop the others. This effectively caps meaningful oracle updates at one per slot regardless of technical capability. Jito BAM creates a marketplace for transaction ordering with verifiable sequencing guarantees, preventing validators from arbitrarily reordering or dropping transactions. This enables multiple oracle updates within a single slot to be processed fairly and predictably.

What are prop AMMs and why do they matter for price discovery?

Prop AMMs (proprietary automated market makers) are a new generation of trading venues where market makers push their own price quotes every slot rather than relying on algorithmic pricing from pool ratios. Traditional AMMs discover prices passively through supply and demand in liquidity pools, while prop AMMs reflect active quotes from professional traders. This dramatically improves price accuracy and reduces spreads. More importantly, with prop AMMs dominating spot trading volume, price discovery is moving on-chain for the first time—meaning traders may eventually look to Solana for reference prices before checking centralized exchanges.

How efficient are modern oracle updates compared to earlier versions?

Switchboard has optimized oracle updates to approximately 20 compute units on-chain—a dramatic improvement from early implementations. However, when market makers push updates every slot for prop AMMs, consumption climbs to 70,000-120,000 compute units per asset. With millions of updates per epoch needed to support just a few prop AMMs, the industry needed a more efficient approach. The new Jito BAM integration allows oracle updates to be inserted immediately before trades execute, eliminating wasteful updates and ensuring resources are spent only when price data is actually needed.

Can Solana compete with traditional exchanges on latency?

The New York Stock Exchange operates at approximately 50-microsecond latency—an extremely high bar for any decentralized system. Until recently, blockchain trading couldn't come close. However, with Jito BAM enabling intra-slot price updates, Solana can now deliver price accuracy at any point within a 200-millisecond slot rather than being limited to slot boundaries. Combined with networking optimizations from solutions like Double Zero that reduce inter-validator latency, the gap is closing significantly. Dr. Blox suggested that microsecond-level pricing is now achievable, putting decentralized venues in direct competition with centralized infrastructure for the first time.