The Real Hail Mary Project on the Moon

Lunar infrastructure is moving out of the conceptual phase and into early engineering definition. Heavy landers, surface power systems, mobility assets, excavation concepts, and permanent outpost discussions are now part of active technical planning. The language of the sector reflects confidence and cadense: deployment, scale-up, operations, utilization. The ground, however, has not advanced at the same pace.

That mismatch deserves direct examination.

From a geotechnical standpoint, the key question is not whether lunar infrastructure is ambitious. Ambition is expected. The real question is more fundamental: what would qualify as a true Hail Mary condition on the Moon? Not in the casual sense of a bold attempt, but in the engineering sense of proceeding with load-bearing systems under boundary conditions that remain insufficiently constrained.

This matters because the Moon is approaching a dangerous threshold. Programs are increasingly framed around architecture, payload mass, energy supply, survivability, and logistics, while the supporting ground is still treated as a secondary surface condition rather than a governing design medium. That is not a semantic problem. It is a project-definition problem. On Earth, no serious underground or surface infrastructure program moves forward without a site investigation strategy, a ground model, parameter ranges, and a clear understanding of how uncertainty will affect performance. On the Moon, the industry is still relying on sparse legacy measurements, orbital interpretation, and simulant-based extrapolations to support decisions that are moving steadily toward infrastructure scale.

The issue is not that the available lunar datasets lack value. They are essential. The issue is that they do not yet provide the kind of geotechnical continuity required for engineering-grade prediction. Discrete observations are being asked to support continuous design logic. That gap is where risk begins to change character. It stops being bounded engineering uncertainty and starts becoming assumption-driven performance.

This is where the Hail Mary concept becomes useful.

A real Hail Mary project on the Moon would not be defined by a specific lander, program, or agency. It would be defined by a condition: advancing infrastructure-scale loading, construction, or sustained operations on ground that has not been characterized at the scale of the demand being imposed on it. In other words, the project stops depending primarily on measured behavior and starts depending on the hope that the regolith will respond within acceptable limits.

That condition is more relevant to lunar development than many current discussions admit. Regolith is still too often simplified as loose granular cover, when the engineering reality is more complex. Lunar ground response is controlled by fabric, density state, stress history, confinement level, and scale effect under an environment that has no true terrestrial equivalent. This is not a detail to be handled later in the design sequence. It is one of the central constraints on whether lunar systems remain stable, serviceable, and scalable.

There is also a business dimension that should not be ignored. Ground uncertainty is not only a technical risk. It is a capital risk, a schedule risk, and a credibility risk. The farther lunar programs move toward infrastructure without a defensible geotechnical basis, the more exposed they become to performance surprises that cannot be cheaply corrected. In terrestrial practice, unresolved ground conditions routinely control contingency, constructability, claims exposure, and long-term serviceability. There is no reason to assume the Moon will be more forgiving. It is more likely to be less forgiving.

This article examines the question directly: what constitutes a real Hail Mary project on the Moon when viewed through the lens of geotechnical engineering? The objective is not rhetorical criticism. It is to establish a practical engineering framework for recognizing when lunar infrastructure crosses the line from managed uncertainty into unbounded ground risk. Once that line is visible, the path forward also becomes clearer: better ground intelligence, more disciplined characterization, and design logic that treats the lunar surface not as background, but as the first-order system it has always been.

From Exploration to Construction Without Transition

The lunar program has moved into a new phase, but the underlying engineering approach has not transitioned with it.

For decades, lunar activity was driven by exploration. The objective was to observe, sample, and interpret. Data density was low, spatial coverage was limited, and the consequences of ground interaction were secondary. A lander footprint, a rover track, or a core sample provided insight, not design constraints.

That context has changed.

Current programs are no longer limited to observation. They are defining:

• Repeated landings at selected sites

• Surface operations with increasing mass and duration

• Fixed assets such as power systems and landing pads

• Early-stage excavation and material handing concepts

It is important to consider that these are not exploration activities. They are infrastructure precursors. In terrestrial practice, this shift triggers a fundamental change in workflow.

The project moves from:

1. Geological Interpretation: understanding what is present

   to

2. Engineering characterization: defining how it will behave under load

That transition is formal, and it should introduce site investigation campaigns, ground models with stratigraphic continuity, parameter ranges for strength, stiffness, and variability, and defined uncertainty classes and design envelops

These elements should serve as the basis on which infrastructure becomes predictable.

On the Moon, this transition has not been formalized. The industry is still operating largely within an exploration-derived framework for reconnaissance such as the data is interpreted globally rather than resolved locally, surface properties are inferred from orbital proxies, mechanical behavior is extrapolated from limited in-situ observations and laboratory simulants.

They are not sufficient for design.

The consequence is subtle but critical. Engineering decisions are being made using datasets that were not generated for engineering purposes. The resolution, scale, and type of information required to predict ground response under load are not aligned with the decisions now being taken.

This creates a structural gap between what is known and what is required.

That gap becomes more pronounced as system scale increases. A single landing event can tolerate uncertainty. Repeated landings at the same site cannot. A rover can adapt to local variability. A fixed foundation cannot. An exploratory trench can accept localized instability. A production excavation cannot.

Without a transition to engineering-grade characterization, each step toward infrastructure amplifies the exposure to ground-driven effects that are not fully bounded.

This is where the Hail Mary condition begins to take shape—not as a single decision, but as a cumulative outcome. The program advances, the systems scale, and the loads increase, while the ground model remains at an exploratory level of resolution.

I see that from a geotechnical standpoint, that is not a neutral trajectory. It is a misalignment between project phase and ground knowledge.

Defining a Hail Mary in Engineering Terms

A Hail Mary project is not characterized by ambition or technical complexity. It is characterized by how decisions are made relative to the available ground information.

From a geotechnical standpoint, the condition exists when three elements are present:

1. Boundary conditions are not bounded: Ground behavior is not defined within measurable limits. Key parameters such as strength, stiffness, and variability are inferred rather than established.

2. Design is based on assumed representativeness: Average values or generalized interpretations are used in place of parameter ranges and envelopes. Local variability is not explicitly incorporated into design logic.

3. Failure modes are only partially defined: The system is evaluated against expected performance, but not against the full range of plausible ground-driven responses under operational conditions.

This is distinct from conventional engineering risk. Also, in standard practice, uncertainty is managed through parameter ranges rather than single values, conservative envelopes rather than best estimates and explicit consideration of variability and scale.

In other words, a Hail Mary condition emerges when uncertainty is no longer contained.

We can assume that the distinction can be expressed clearly:

Aspects	Managed Engineering Condition	Hail Mary Condition
Ground Model	Continuous, parameterized	Discrete, inferred
Input Parameter	Ranges and bonds	Representative values
Variability	Explicitly considered	Implicit or ignored
Verification	Field-based or staged	Limited or deferred
Design Outcome	Predicable within limits	Dependent on actual ground response

In the context, the issue is not whether the design is technically sound. A system can be well designed relative to its inputs and still fall into a Hail Mary condition if those inputs do not adequately represent the ground. The concept also has a temporal dimension. Early-stage concepts often operate with limited data, which is acceptable. The condition becomes critical when:

• System definition advances

• Loads increase

• Operational dependency on ground performance grows.

At that point, the project is no longer exploratory in nature, but the ground model remains exploratory in resolution. This is the threshold where engineering transitions into a Hail Mary condition.

In the lunar context, this definition provides a practical lens. It allows current and future infrastructure concepts to be evaluated not on their ambition, but on whether the ground behavior they depend on has been sufficiently bounded to support the decisions being made.

What "Non-Hail Mary" Looks Like in the Terrestrial Baseline

Before assessing the Moon, the reference condition must be clear. In terrestrial geotechnical engineering, infrastructure does not proceed on the basis of assumed ground behavior. It proceeds on the basis of bounded ground response.

That boundary is not achieved through a single dataset. It is built through a structured sequence that links investigation, interpretation, and design.

At minimum, this sequence includes:

• Site investigation aligned with project scale

  Boreholes, in-situ testing, and geophysics are planned to capture variability at the scale of the structure, not just the site.

• Stratigraphic and geological model

  Ground is represented as a continuous system with defined units, interfaces, and spatial relationships.

• Parameterization of behavior

  Strength, stiffness, deformability, and permeability are defined as ranges, not single values, with attention to anisotropy and disturbance.

• Stress condition and history

  In-situ stress state and stress history are evaluated, as they directly control deformation and failure mechanisms.

• Design envelopes and verification strategy

  Loads are checked against parameter ranges, and the design is supported by staged verification through instrumentation and monitoring.

This is pretty much a standard practice in the construction industry, and the outcome, of course, is not perfect prediction. It is predictability within defined limits.

However, the distinction is purely operational, then, when ground response is bounded you should expect that settlement can be estimated with a range, bearing capacity can be checked against limit states, and construction methods can be selected with known performance expectations.

Therefore, when ground response is not bounded, the same checks become conditional.

The difference can be summarized as follow:

Element	Bounded Condition (Non-Hail Mary)	Unbounded Condition
Stratigraphy	Continuous and correlated	Discrete and inferred
Parameters	Defined ranges	Representative values
Stress state	Measured or constrained	Assumed
Variability	Quantified	Qualitative
Design checks	Limit states verified	Performance assumed

A critical aspect of this framework is scale consistency. The level of investigation is matched to the level of demand. A shallow foundation supporting low loads requires one level of resolution. A deep excavation or high-capacity structure requires another. The investigation evolves with the project.

Equally important is the feedback loop. Design is not static. Instrumentation, monitoring, and observational methods are used to confirm or adjust the ground model as construction progresses. This reduces uncertainty in real time.

None of these elements eliminate risk. They convert it into a form that can be managed, priced, scheduled, and controlled.

This baseline matters because it defines the threshold. A project does not become a Hail Mary because it is difficult or innovative. It becomes a Hail Mary when it bypasses, compresses, or defers this sequence while still depending on the ground to perform within acceptable limits.

Finally, with that reference established, and clear in your mind, the lunar condition can be evaluated directly, not against expectations, but against the standard that governs every serious geotechnical project on Earth.

Applying the Criteria to the Moon. Where the Gaps Emerges?

With the terrestrial baseline established, the lunar condition can be evaluated directly. The question is not whether data exists. The question is whether the available information satisfies the requirements for bounded ground behavior under load.

At present, it does not.

The gap is not singular. It appears across multiple layers of the ground model, each of which contributes to how regolith will respond under infrastructure-scale loading.

The Absence of a Continuous Ground Model

Lunar datasets are inherently discontinuous. Apollo measurements are spatially isolated, and orbital data provides surface-scale interpretation. Additionally, and more importantly, simulant testing offers controlled but simplified behavior.

What is missing is correlation across scale.

There is no established stratigraphic continuity at the resolution required for engineering design. Interfaces, layering, and lateral variability are inferred rather than mapped. This limits the ability to define how loads will transfer through the subsurface.

In other words, the lunar ground is known in localized points but not in volume.

Stress History Is Not Characterized

The mechanical behavior of lunar regolith is governed by its stress history.

Impact processes, repeated disturbance, and long-term densification produce a material that does not follow the assumptions typically associated with loose granular soils. Yet this stress-history dimension is not currently parameterized in design workflows.

This is where the absence of a framework such as OCR* becomes critical.

Without a means to quantify:

• Degree of overconsolidation

• Depth-dependent densification

• Sensitivity to disturbance

The transition between apparent stiffness and deformation cannot be bounded. Designs implicitly assume behavior that has not been verified.

Boundary Conditions Are Non-Terrestrial

It is extremely important to highlight that Lunar ground response is controlled by conditions that have no direct terrestrial equivalent such as extremely low confinement due to reduced gravity, vacuum environment affecting particle interaction, electrostatic forces influencing surface behavior, and thermal cycling altering near-surface fabric.

These factors influence how regolith mobilizes strength and undergoes deformation. Laboratory simulants can approximate some aspects, but they do not fully reproduce the combined effect of these conditions at scale.

As a result, the governing boundary conditions of the problem remain only partially represented in current datasets.

Scale Mismatch Between Data and Demand

The available data reflects small-scale interaction. However, the future system will introduce higher loads, larger contact cycles, repeated loading cycles and longer operational durations, which haven't been properly evaluated and accounted.

This creates a scale mismatch.

Aspect	Available Data Scale	Infrastructure Demand Scale
Load interaction	Small, localized	Large, distributed
Depth of influence	Shallow	Deeper, cumulative
Duration	Short-term	Repeated and sustained
Variability capture	Limited	Operationally critical

Extrapolating from one scale to another without intermediate validation introduces uncertainty that cannot be easily bounded.

Variability Is Not Quantified

Variability is central to geotechnical design. On Earth, even well-characterized sites include parameter ranges to account for spatial differences. On the Moon, variability is acknowledged qualitatively but not quantified in a way that supports design envelopes.

Key unknowns include:

• lateral variation in density and fabric

• depth-dependent changes in stiffness

• localized zones of weakness or disturbance

Without quantified variability, design checks cannot be performed against worst-case or bounding conditions. They rely instead on assumed representativeness.

Conditions for a Hail Mary Are Present

When evaluated against the criteria defined earlier, the lunar condition aligns with a Hail Mary framework. Here the takeaway:

• Ground behavior is not bounded by measured parameter ranges

• Design inputs rely on inferred or representative values

• Failure modes cannot be fully enumerated across operational scales

This does not imply that current efforts are flawed. It reflects the stage of development. The data that exists was not collected to support infrastructure design at scale. The issue is that system definition is advancing faster than ground characterization. From a geotechnical standpoint, this is the point where uncertainty transitions from managed to unbounded.

What Would a Real Hail Mary Project Look Like?

A real Hail Mary project on the Moon is not defined by a specific mission or system. It is defined by proceeding with infrastructure-scale decisions while the governing ground behavior remains unbounded. This condition can be described in operational terms.

A project enters a Hail Mary state when it advances under the following characteristics:

Design Element	Hail Mary Condition in Lunar Context
Bearing capacity	Not derived from site-specific parameters or load tests
Settlement	Not bounded under static or repeated loading conditions
Ground model	Based on discrete observations without continuity
Variability	Recognized but not quantified in design inputs
Stress history (OCR*)	Not characterized or incorporated
Scale validation	No intermediate testing between small-scale data and full-scale demand
Verification	Deferred or absent prior to operational deployment
Compaction	Not quantified under cyclic loading, traffic and repeated load application

Under these conditions, performance is not predicted within limits. It is assumed to fall within acceptable behavior.

Where This Condition Manifests

As system concepts move from single interactions to repeated and sustained use of the surface, the dependence on ground performance increases, while the underlying characterization remains limited. The result is a set of operations that are technically feasible but not yet bounded in geotechnical terms.

The manifestation can be understood through typical use cases:

Operation Type	Ground Dependency	Unbounded Aspect
Repeated heavy landings	Load applied cyclically to the same footprint	Cumulative compaction, densification, and settlement not quantified
Permanent surface assets	Continuous load over time	Long-term settlement and bearing behavior not bounded
Landing pad development	Surface modification for load distribution	Subsurface variability and response not characterized
Excavation and material handling	Disturbance of in-situ fabric	Stability governed by stress history and structure not defined
Mobility corridors under repeated traffic	Progressive surface loading	Degradation, rutting, and densification not constrained

In each case, the engineering challenge is not feasibility. The systems can be built and deployed. The issue is that ground response is not defined within limits that allow prediction of performance over time.

The transition from isolated events to repeated operations changes the problem fundamentally.

A single landing can tolerate local variability. The system interacts with the ground once, and the consequence is contained. When that same location is used repeatedly, the ground evolves, fabric changes, density increases locally, and deformation accumulates. The problem becomes path dependent.

The same applies across operations. A rover moving across undisturbed terrain responds to existing conditions. A trafficked corridor modifies those conditions with every pass. An exploratory excavation can rely on observational adjustment. A production excavation depends on predictable behavior across its extent.

This distinction defines the threshold.

It is not the magnitude of an individual load that governs performance. It is the interaction between load, repetition, and ground evolution. Once operations rely on repeated interaction, the absence of bounded parameters becomes critical.

What Is Being Assumed?

At the core of this condition is a set of assumptions that are currently necessary but not yet supported by sufficient data at scale.

Assumed Behavior	Engineering Requirement	Current Status
Bearing capacity remains adequate	Site-specific strength parameters and variability	Not established
Settlement remains within serviceability limits	Bounded response under static and cyclic loading	Not quantified
Compaction and densification remain controlled	Response under repeated loading and disturbance	Not characterized
Regolith behavior is sufficiently uniform	Spatial variability defined across footprint	Not resolved
Disturbance does not alter performance significantly	Evolution of fabric and stiffness understood	Not bounded

These assumptions are not flawed. They are unavoidable at the current stage. The issue is that they remain open-ended.

In engineering terms, the system is relying on conditions that have not yet been constrained into design parameters.

Answering the Question

A real Hail Mary project on the Moon is therefore defined by a single condition:

Proceeding with infrastructure-scale interaction with the ground while the governing parameters of that interaction remain insufficiently constrained.

This is not a matter of intent or design quality. It is a matter of alignment between:

• the demands imposed by the system, and

• the level of ground knowledge available to support those demands

When that alignment is missing, performance depends on the ground behaving within acceptable limits after deployment, rather than being predicted within those limits before it.

From a geotechnical standpoint, that is the Hail Mary condition.

The Real Hail Mary

The question is not whether lunar infrastructure is feasible. It is whether it is being defined on a geotechnical basis that can support its intended performance.

From that standpoint, the real Hail Mary project on the Moon is now clear.

It is not a specific mission, vehicle, or architecture. It is the condition of advancing infrastructure-scale interaction with the ground while the governing parameters of that interaction remain insufficiently constrained.

This condition is not unusual at early stages. Exploration programs operate with limited data by definition. What is changing is the level of demand placed on that data. Systems are increasing in mass, repetition, and operational dependency on the ground, while the underlying characterization remains at an exploratory level of resolution.

That misalignment is where risk changes form.

On Earth, this transition is managed through a structured process: investigation, parameterization, verification, and continuous refinement of the ground model. This process does not eliminate uncertainty, but it defines it. It allows design to proceed within limits that are understood, monitored, and adjusted as needed.

On the Moon, that process has not yet been fully established at the program level.

As a result, several key aspects of ground behavior remain open:

• bearing capacity at operational scale

• settlement under repeated loading

• compaction and densification effects over time

• variability across the footprint of infrastructure

• the role of stress history in governing response

The implication is direct. The constraint on lunar infrastructure is not only technological. It is geotechnical.

Moving out of the Hail Mary condition does not require a fundamental change in ambition. It requires a change in sequence. Ground characterization must advance in parallel with system definition, not behind it. The focus shifts from assuming acceptable performance to establishing the parameters that define it.

This is where frameworks that capture regolith behavior in engineering terms become necessary. Stress history, expressed through parameters such as OCR*, provides a path to move from qualitative understanding to quantitative design. Combined with structured classification and suitability metrics, it allows ground response to be incorporated into decision-making rather than treated as a boundary condition to be verified after deployment.

The path forward is consistent with established engineering practice:

• targeted in-situ investigation

• scale-appropriate testing

• parameter-based design envelopes

• staged validation and feedback

None of these elements are new. What is new is the environment in which they must be applied.

The Moon will not respond to infrastructure differently because it is a new domain. It will respond according to its ground conditions. Those conditions must be defined, bounded, and integrated into design.

Until that occurs, any infrastructure system, regardless of its level of technological sophistication, remains dependent on an assumption:

That the ground will behave within acceptable limits.

From a geotechnical perspective, that is the real Hail Mary.

Roberto de Moraes

Author |SpaceGeotech Founder